U.S. patent application number 13/811904 was filed with the patent office on 2013-09-26 for touch sensitive device.
This patent application is currently assigned to HIWAVE TECHNOLOGIES (UK) LIMITED. The applicant listed for this patent is Neil Harris. Invention is credited to Neil Harris.
Application Number | 20130249831 13/811904 |
Document ID | / |
Family ID | 42752691 |
Filed Date | 2013-09-26 |
United States Patent
Application |
20130249831 |
Kind Code |
A1 |
Harris; Neil |
September 26, 2013 |
TOUCH SENSITIVE DEVICE
Abstract
A method of generating a set of filters comprising: determining
an initial estimate of a filter to be applied to a signal
associated with each transducer; defining a model of the system;
calculating the vibration of the member as an output of the model
of the system; calculating the output of the model by comparing the
output of the model with a measured value; determining changed
parameter values of parameters of the model; recalculating the
error value for the output of the model by comparing the output of
the model with the changed parameter values with the measured
value; comparing the recalculated error value with the reference
error value; setting the recalculated error value as the reference
error value, setting the changed parameter values as the model
parameters, and repeating the above steps, or outputting the model
parameters; generating a set of new filters each using respective
output model parameters.
Inventors: |
Harris; Neil; (Cambourne,
GB) |
|
Applicant: |
Name |
City |
State |
Country |
Type |
Harris; Neil |
Cambourne |
|
GB |
|
|
Assignee: |
HIWAVE TECHNOLOGIES (UK)
LIMITED
Cambridgeshire
GB
|
Family ID: |
42752691 |
Appl. No.: |
13/811904 |
Filed: |
July 25, 2011 |
PCT Filed: |
July 25, 2011 |
PCT NO: |
PCT/GB2011/051417 |
371 Date: |
June 11, 2013 |
Current U.S.
Class: |
345/173 |
Current CPC
Class: |
G06F 3/041 20130101;
G08B 6/00 20130101; G06F 3/0436 20130101; G06F 3/016 20130101 |
Class at
Publication: |
345/173 |
International
Class: |
G06F 3/041 20060101
G06F003/041 |
Foreign Application Data
Date |
Code |
Application Number |
Jul 23, 2010 |
GB |
1012387.5 |
Claims
1. A method of generating a set of filters for a touch sensitive
device comprising a touch-sensitive member and a plurality of
transducers mounted to the member, the method comprising: a)
determining an initial estimate of a filter to be applied to a
respective signal associated with each transducer; b) defining a
model of the system whereby the relationship of vibration of the
member to the respective signals can be calculated, the model
having a plurality of parameters; c) calculating the output of the
model of the system; d) calculating a reference error value for the
output of the model by comparing the output of the model with a
measured value; e) determining changed parameter values of the
parameters of the model; f) recalculating the error value for the
output of the model by comparing the output of the model with the
changed parameter values with the measured value; g) comparing the
recalculated error value with the reference error value; h) if the
compared recalculated error value is less than the reference error
value, setting the recalculated error value as the reference error
value, setting the changed parameter values as the model
parameters, and repeating the steps c) to h), or if the compared
recalculated error value is greater than the reference error value,
outputting the model parameters; generating a set of new filters
each using respective output model parameters.
2. The method according to claim 1, wherein the filter has a
plurality of coefficients, the number of filter coefficients being
equal to the number of model parameters.
3. The method according to claim 1, wherein the model comprises an
inverse of the filter.
4. The method according to claim 1, wherein the reference error
value is calculated using: SSE ( a , b , ) = n Y ( n , a , b , ) -
X ( n ) 2 ##EQU00041## where X(n) is the measured value which is
measured at n test points and Y(n) is the output of the system
model where a, b are the parameter values.
5. The method according to claim 4, wherein determining changed
parameter values of the parameters of the model comprises selecting
parameters to reduce the value of SSE by finding the value of t
that minimizes F(t) where F ( t ) = SSE ( a + t .differential. SSE
.differential. a , b + t .differential. SSE .differential. b , )
##EQU00042##
6. The method according to claim 5, comprising finding the value of
t by estimating the vector gradient of error (grad (SSE)), changing
all the model parameters proportionally to the estimated vector
gradient of error, estimating F'(t)=0 and F''(t)=0 and determining
t from the estimated derivatives of F(t) where grad ( SSE ) = (
.differential. SSE .differential. a , .differential. SSE
.differential. b , ) T ##EQU00043## and ##EQU00043.2## F ' ( t ) =
.differential. F .differential. t and F '' ( t ) = .differential. 2
F .differential. t 2 . ##EQU00043.3##
7. The method according to claim 6, comprising estimating the
vector gradient of error with respect to each of the parameters by
changing each parameter by a small amount, recalculating SSE and
estimating the vector gradient of error using the difference
between the reference value for SSE and the recalculated value of
SSE.
8. The method according to claim 6, comprising estimating F'(t=0)
and F''(t=0) by determining two new parameter sets, the first
parameter set being changed proportionally to the estimated vector
gradient of error by addition and the second parameter set being
changed proportionally to the estimated vector gradient of error by
addition, calculating SSE values for each of the two new parameter
sets and calculating estimates for F'(t=0) and F''(t=0) from the
new SSE values.
9. A method of generating a desired touch sensitivity in a touch
sensitive device comprising a touch-sensitive member and a
plurality of transducers mounted to the member, the method
comprising: generating a set of filters by carrying out the method
of claim 1; applying the set of filters to an output signal from
each transducer to generate filtered output signals; and using the
filtered output signals to provide the desired touch
sensitivity.
10. The method according to claim 9, wherein: said determining an
initial estimate comprises determining an initial estimate of a
filter to be applied to a respective signal output from each
transducer, whereby the filtered signal is generated by vibration
of the member resulting from a physical touch on the member; the
model of the system is defined whereby the output signal can be
calculated as a function of vibration of the member; said
calculating the output of the model comprises calculating the
output signal as an output of the model of the system; said
calculating a reference error value comprises calculating a
reference error value for the output of the model by comparing the
output of the model with a measured value of the output signal when
a known touch is applied to the member; and said recalculating the
error value comprises recalculating the error value for the output
of the model by comparing the output of the model with a measured
value of the output signal when a known touch is applied to the
member.
11. The method according to claim 1, wherein the reference error
value is calculated using: SSE = chan Q chan * h chan 2 chan L chan
* h chan 2 ##EQU00044## where the star signifies the convolution
product, h.sub.chan is the time impulse response for the filtered
output signal at each transducer, Q.sub.chan is the output produced
by each transducer at a relatively touch insensitive location and
L.sub.chan is the output produced by each transducer at a
relatively touch sensitive location.
12. The method according to claim 11, wherein recalculating the
error value comprises perturbing each time impulse response h by a
small amount .alpha. at each transducer and calculating a new error
value from SSE ( .alpha. ) = chan Q chan * ( h chan + .alpha.
.delta. chan , j z - n .delta. ( 0 ) ) 2 chan L chan * ( h chan +
.alpha. .delta. chan , j z - n .delta. ( 0 ) ) 2 ##EQU00045## Where
z is the unit delay operator .delta. is the delta (sampling)
function and .delta..sub.i,j is the Kronecker delta.
13. The method according to claim 12, wherein recalculating the
error value comprises defining the gradient vector as; follows:
grad ( SSE ) n , j = lim a -> 0 ( SSE ( .alpha. ) - SE .alpha. )
##EQU00046## grad ( SSE ) n , j = 2 chan L chan * h chan 2 ( ( z -
n Q j ) ( chan Q chan * h chan ) - SSE ( z - n L j ) ( chan L chan
* h chan ) ) ##EQU00046.2## and solving to find an exact value for
the gradient vector.
14. The method according to claim 13, wherein recalculating the
error value comprises determining the value at which the gradient
vector is zero and recalculating the error value for this
value.
15. The method according to claim 1, comprising determining the
initial filter as a time-reversal filter.
16. The method according to claim 1, comprising determining the
initial filter as a simultaneous multi-region filter.
17. The method according to claim 2, comprising determining the
initial estimate using arbitrary filter coefficients.
18. The method according to claim 1, comprising determining the
initial estimate using an infinite impulse response filter.
19. The method according to claim 1, wherein the desired touch
sensitivity is to produce a maximum response to a touch at one
location on the touch sensitive member and a minimum response to a
touch at a second location on the touch sensitive member.
20. A touch sensitive device comprising a touch-sensitive member, a
plurality of transducers mounted to the member, and a processor
configured to carry out the method according to claim 1.
21. A method of generating a set of filters for a touch sensitive
device comprising a touch-sensitive member and a plurality of
transducers mounted to the member, the method comprising: choosing
a set of frequencies for use in the filters; determining an impulse
response of a filter for each respective transducer to be applied
to a signal associated with each transducer; calculating the
transfer function of each filter, wherein each filter has a
transfer function with at least one pole and at least one zero, and
calculating the transfer function of each said filter comprises,
determining at least one pole coefficient which determines at least
one pole; determining, using said at least one pole coefficient, a
pole representation of the transfer function which filters said
input signal using said at least one pole; using said respective
impulse response, calculating at least one zero coefficient which
determines at least one zero; and combining said pole
representation of the transfer function with said at least one zero
coefficient to calculate said transfer function of said filter; and
generating a set of filters comprising said calculated filters.
22. The method according to claim 21 wherein each said filter is an
infinite impulse response filter.
23. The method according to claim 22 wherein said infinite impulse
response filter has a transfer function of the form: H z ( z , d )
:= k ( d k , 0 + d k , 1 z - 1 + d k , 2 z - 2 1 + a k , 0 z - 1 +
a k , 1 z - 2 ) ##EQU00047## where d.sub.k,0 d.sub.k,1 d.sub.k,2
are zero coefficients which determine the zeros, a.sub.k,0
a.sub.k,1 a.sub.k,2 are pole coefficients which determine the poles
and k is the number of poles.
24. The method according to claim 23, wherein the pole coefficients
are expressed as a.sub.k,0=-2Re(p.sub.k) and
a.sub.k,1=|p.sub.k|.sup.2 and the transfer function is written as H
z ( z , d ) = k d k , 0 + d k , 1 z - 1 + d k , 2 z - 2 ( 1 - p k z
- 1 ) ( 1 - p _ k z - 1 ) ##EQU00048##
25. The method according to claim 23 or claim 21, comprising
determining said pole representation of the transfer function from:
u.sub.0,k:=1 u.sub.1,k:=-(a.sub.k,0u.sub.0,k)
u.sub.i+2,k:=-(a.sub.k,0u.sub.i+1,k+a.sub.k,1u.sub.i,k)
26. The method according to claim 21, comprising determining the
pole coefficients from the chosen set of frequencies.
27. The method according to claim 21, comprising calculating the
zero coefficients using a generalized inverse matrix with the zeros
being defined by: z 1 k , j := d 3 k + 1 , j - ( d 3 k + 1 , j ) 2
- 4 d 3 k , j d 3 k + 2 , j - 2 d 3 k , j ##EQU00049## z 2 k , j :=
d 3 k + 1 , j + ( d 3 k + 1 , j ) 2 - 4 d 3 k , j d 3 k + 2 , j - 2
d 3 k , j . ##EQU00049.2##
28. The method according to claim 21, comprising calculating zero
coefficients for zeros which are not paired with said determined
pole coefficients.
29. A method of generating a desired touch sensitivity in a touch
sensitive device comprising a touch-sensitive member and a
plurality of transducers mounted to the member, the method
comprising: generating a set of filters by carrying out the method
of claim 21; applying the set of filters to an output signal from
each transducer to generate filtered output signals; and using the
filtered output signals to provide the desired touch
sensitivity.
30. The method according to claim 29, wherein: said determining an
impulse response of the filter comprises determining an impulse
response of the filter for each respective transducer to be applied
to a signal output by said transducer, wherein the output signal is
generated by vibration of the member resulting from a physical
touch on the member.
31. The method according to claim 21 wherein the impulse response
of each filter is derived from a measured value of vibration of the
member.
32. A touch sensitive device comprising a touch-sensitive member, a
plurality of transducers mounted to the member, and a processor
configured to carry out the method according to claim 21.
33. A method of generating a set of filters for a touch sensitive
device comprising a touch-sensitive member and a plurality of
transducers mounted to the member, the method comprising: choosing
a set of frequencies for use in the filters; calculating a set of
transfer functions of respective filters for each transducer to be
applied to a signal associated with each transducer; wherein each
filter has a transfer function with at least one pole and at least
one zero and calculating the transfer function of each said filter
comprises, determining at least one pole coefficient which
determines said at least one pole; determining, using said at least
one pole coefficient, a pole representation of the transfer
function which filters said input signal using said at least one
pole; using an eigenvector method to determine at least one zero
coefficient which determines said at least one zero; and combining
said pole representation of the transfer function with said at
least one zero coefficient to determine said transfer function for
said filter; and generating a set off filters comprising said
calculated filters.
34. The method according to claim 33 wherein said filter is an
infinite impulse response filter.
35. The method according to claim 33, further comprising
determining zero coefficients for zeros which are not paired with
said determined pole coefficients.
36. A method of generating a desired touch sensitivity in a touch
sensitive device comprising a touch-sensitive member and a
plurality of transducers mounted to the member, the method
comprising: generating a set of filters by carrying out the method
of claim 33; applying the set of filters to an output signal from
each transducer to generate filtered output signals; and using the
filtered output signals to provide the desired touch
sensitivity.
37. The method according to claim 36, wherein: said calculating a
set of transfer functions of respective filters for each transducer
comprises calculating a set of transfer functions of respective
filters for each transducer to be applied to a signal output by
each transducer, wherein the output signal is generated by
vibration of the member resulting from a physical touch on the
member.
38. The method of claim 37 wherein the eigenvector method
comprises: defining a first parameter of the desired touch
sensitivity; using an eigenvector method to identify sets of
possible filter transfer functions which satisfy the first
parameter; defining a second parameter of the desired touch
sensitivity; using an eigenvector method to select from the
identified sets of possible filter transfer functions which satisfy
the first parameter, sets of possible filter transfer functions
which satisfy the second parameter.
39. The method of claim 38 wherein the eigenvector method further
comprises: defining at least one further parameter of the desired
touch sensitivity; and using an eigenvector method to select from
the previously selected sets of possible filter transfer functions
which satisfy the first parameter and second parameter, sets of
possible filter transfer functions which best satisfy the at least
one further parameter.
40. The method of claim 39 wherein the selection selects a
combination of possible filter transfer functions.
41. The method of claim 40 wherein the selection selects a linear
combination of possible filter transfer functions.
42. A touch sensitive device comprising a touch-sensitive member, a
plurality of transducers mounted to the member, and a processor
configured to carry out the method according to claim 33.
43. A computer program comprising program code which, when executed
on a processor of a touch sensitive device, will cause the touch
sensitive device to carry out the method claim 1.
Description
TECHNICAL FIELD
[0001] The invention relates to touch sensitive devices including
touch sensitive screens or panels.
BACKGROUND ART
[0002] U.S. Pat. No. 4,885,565, U.S. Pat. No. 5,638,060, U.S. Pat.
No. 5,977,867, US2002/0075135 describe touch-operated
apparatus.
DISCLOSURE OF INVENTION
[0003] According to a first aspect of the invention, there is
provided a method of generating a set of filters for a touch
sensitive device comprising a touch-sensitive member and a
plurality of transducers mounted to the member, the method
comprising: [0004] a) determining an initial estimate of a filter
to be applied to a respective signal associated with each
transducer; [0005] b) defining a model of the system whereby the
relationship of vibration of the member to the respective signals
can be calculated, the model having a plurality of parameters;
[0006] c) calculating the output of the model of the system; [0007]
d) calculating a reference error value for the output of the model
by comparing the output of the model with a measured value; [0008]
e) determining changed parameter values of the parameters of the
model; [0009] f) recalculating the error value for the output of
the model by comparing the output of the model with the changed
parameter values with the measured value; [0010] g) comparing the
recalculated error value with the reference error value; [0011] h)
if the compared recalculated error value is less than the reference
error value, setting the recalculated error value as the reference
error value, setting the changed parameter values as the model
parameters, and repeating the steps c) to h), or
[0012] if the compared recalculated error value is greater than the
reference error value, outputting the model parameters; generating
a set of new filters each using respective output model
parameters.
[0013] According to a second aspect of the invention, there is
provided a method of generating a desired touch sensitivity in a
touch sensitive device comprising a touch-sensitive member and a
plurality of transducers mounted to the member, the method
comprising:
[0014] generating a set of filters by carrying out the method of
the first aspect;
[0015] applying the set of filters to an output signal from each
transducer to generate filtered output signals; and
using the filtered output signals to provide the desired touch
sensitivity.
[0016] The filter may have a plurality of coefficients, the number
of filter coefficients being equal to the number of model
parameters.
[0017] The initial estimate of the filter to be applied to a signal
output from each transducer, may be an estimate of the required
filter. The initial estimate of the filter to be applied to a
signal output from each transducer, may be any initial estimate of
the filter because the reference error minimisation procedure of
the above method will determine the required output model
parameters. In some examples the initial estimate of the filter may
be a standard, or default, filter, or even a random filter. In some
examples the initial estimate of the filter may be a close estimate
of the required filter. This may provide the advantage of reducing
the number of iterations of the method required to arrive at a
filter giving acceptable performance.
[0018] The above method models the filter in the time domain and
applies an iterative refinement algorithm (repeating changing,
recalculating and comparing steps), to improve the performance of
the filter. The initial filter may be a time-reversal (TR) filter,
a simultaneous multi-region (SMR) filter or an infinite impulse
response filter. SMR filters obtained analytically are exact in the
frequency domain but seldom achieve a good separation in the time
domain. In some examples the application of the refinement
algorithm may provide double the separation. The model may comprise
an inverse of the filter.
[0019] The initial filter may be very complex in the temporal
domain. However, in the frequency domain there may be one or more
key or pole frequencies that are more important than other
frequencies across the frequency range of interest. These one or
more key or pole frequencies may be transformed into separate
initial filter components in the temporal domain. A less complex
initial filter may be represented by the combination of these
separate initial filter components.
[0020] According to a third aspect of the invention, there is
provided a method of generating a set of filters for a touch
sensitive device comprising a touch-sensitive member and a
plurality of transducers mounted to the member, the method
comprising: [0021] choosing a set of frequencies for use in the
filters; [0022] determining an impulse response of a filter for
each respective transducer to be applied to a signal associated
with each transducer; [0023] calculating the transfer function of
each filter, wherein each filter has a transfer function with at
least one pole and at least one zero, and calculating the transfer
function of each said filter comprises, [0024] determining at least
one pole coefficient which determines at least one pole; [0025]
determining, using said at least one pole coefficient, a pole
representation of the transfer function which filters said input
signal using said at least one pole; [0026] using said respective
impulse response, calculating at least one zero coefficient which
determines at least one zero; and [0027] combining said pole
representation of the transfer function with said at least one zero
coefficient to calculate said transfer function of said filter; and
[0028] generating a set of filters comprising said calculated
filters.
[0029] According to a fourth aspect of the invention, there is
provided a method of generating a desired touch sensitivity in a
touch sensitive device comprising a touch-sensitive member and a
plurality of transducers mounted to the member, the method
comprising: [0030] generating a set of filters by carrying out the
method of the third aspect; [0031] applying the set of filters to
an output signal from each transducer to generate filtered output
signals; and [0032] using the filtered output signals to provide
the desired touch sensitivity.
[0033] The impulse response of each filter may be derived from a
desired value of touch sensitivity of the member.
[0034] The filter may be an infinite impulse response filter and
may have a transfer function of the form:
H z ( z , d ) := k ( d k , 0 + d k , 1 z - 1 + d k , 2 z - 2 1 + a
k , 0 z - 1 + a k , 1 z - 2 ) ##EQU00001##
Where d.sub.k,0 d.sub.k,1 d.sub.k,2 are zero coefficients which
determine the zeros, a.sub.k,0 a.sub.k,1 a.sub.k,2 are pole
coefficients which determine the poles and k is the number of
poles.
[0035] The pole coefficients may be expressed as
a.sub.k,0=-2Re(p.sub.k) and a.sub.k,1=|p.sub.k|.sup.2
and the transfer function may be written as
H z ( z , d ) = k d k , 0 + d k , 1 z - 1 + d k , 2 z - 2 ( 1 - p k
z - 1 ) ( 1 - p _ k z - 1 ) ##EQU00002##
[0036] Determining said pole representation of the transfer
function may comprise determining:
u.sub.0,k:=1 u.sub.1,k:=-(a.sub.k,0u.sub.0,k)
u.sub.i+2,k:=-(a.sub.k,0u.sub.i+1,k+a.sub.k,1u.sub.i,k)+
[0037] The pole coefficients may be determined from the chosen set
of frequencies.
[0038] If the output signal for each transducer is known, the zero
coefficients may be determined from
z 1 k , j := d 3 k + 1 , j - ( d 3 k + 1 , j ) 2 - 4 d 3 k , j d 3
k + 2 , j - 2 d 3 k , j ##EQU00003## z 2 k , j := d 3 k + 1 , j - (
d 3 k + 1 , j ) 2 - 4 d 3 k , j d 3 k + 2 , j - 2 d 3 k , j
##EQU00003.2##
[0039] The method may further comprise calculating zero
coefficients for zeros which are not paired with said determined
pole coefficients.
[0040] According to a fifth aspect of the invention, there is
provided a method of generating a set of filters for a touch
sensitive device comprising a touch-sensitive member and a
plurality of transducers mounted to the member, the method
comprising: [0041] choosing a set of frequencies for use in the
filters; [0042] calculating a set of transfer functions of
respective filters for each transducer to be applied to a signal
associated with each transducer; [0043] wherein each filter has a
transfer function with at least one pole and at least one zero and
calculating the transfer function of each said filter comprises,
[0044] determining at least one pole coefficient which determines
said at least one pole; [0045] determining, using said at least one
pole coefficient, a pole representation of the transfer function
which filters said input signal using said at least one pole;
[0046] using an eigenvector method to determine at least one zero
coefficient which determines said at least one zero; and [0047]
combining said pole representation of the transfer function with
said at least one zero coefficient to determine said transfer
function for said filter; and generating a set of filters
comprising said calculated filters.
[0048] According to a sixth aspect of the invention, there is
provided a method of generating a desired touch sensitivity in a
touch sensitive device comprising a touch-sensitive member and a
plurality of transducers mounted to the member, the method
comprising: [0049] generating a set of filters by carrying out the
method of the sixth aspect; [0050] applying the set of filters to
an output signal from each transducer to generate filtered output
signals: and using the filtered output signals to provide the
desired touch sensitivity.
[0051] The filter may be an infinite impulse response filter.
[0052] The method may further comprise calculating zero
coefficients for zeros which are not paired with said determined
pole coefficients.
[0053] The following features may apply to all aspects.
[0054] The initial reference error value may be calculated using a
sum squared error, for example from:
SSE ( a , b , ) = n Y ( n , a , b , ) - X ( n ) 2 ##EQU00004##
where X(n) is the measured value which is measured in response to a
touch at n test points and Y(n) is the output of a system model
where a, b are the model parameter values. The model parameter
values may correspond to the filter coefficients.
[0055] The SSE is used as an exemplar method for determining a
measure of error between a measured value or values and modelled
value or values. Other methods such as the variance, standard
deviation, mean squared error, root mean square error or other
techniques as would be appreciated by the person skilled in the
art. The SSE is an example of a suitable method only, the present
claimed Method and Device employ any or all of the methods
described above, including those appreciated by the person skilled
in the art but not explicitly mentioned herein.
[0056] The SSE may be minimised by any suitable minimisation
routine. A number of known minimisation routines may be used such
as Gradient Search or Gradient Decent, Synthetic Annealing, Newton
and Quasi-Newton methods, Interior point methods, linear least
squares methods (a regression model comprising a linear combination
of the parameters), functional analysis methods (approximating to a
sum of other functions), non-linear least squares methods
(approximate to a linear model and refine the parameters by
successive iterations) and other methods that would be appreciated
by the skilled person in the art.
[0057] A Gradient Search method is described below for the
minimisation of the SSE. This is an example of a suitable method
only, the present claimed Method and Device may employ any or all
of the methods described above, including those appreciated by the
person skilled in the art but not explicitly mentioned herein.
[0058] Changing the values of the parameters of the model may
comprise selecting parameters to reduce the value of SSE by finding
the value of t that minimises F(t) where
F ( t ) = SSE ( a + t .differential. SSE .differential. a , b + t
.differential. SSE .differential. b , ) ##EQU00005##
[0059] This means that F(t) can be written as
F ( t ) = n Y ( n , a + t .differential. SSE .differential. a , b +
t .differential. SSE .differential. b , ) - X ( n ) 2
##EQU00006##
[0060] Finding the value of t may comprise estimating the vector
gradient of error (grad (SSE)), changing all the model parameters
proportionally to the estimated vector gradient of error,
estimating the first and second derivatives F'(t=0) and F''(t=0) of
F(t=0) for the value of t equal to zero and determining t from the
estimated derivatives of F(t) where
grad ( SSE ) = ( .differential. SSE .differential. a ,
.differential. SSE .differential. b , ) T and ##EQU00007## F ' ( t
) = .differential. F .differential. t and F '' ( t ) =
.differential. 2 F .differential. t 2 . ##EQU00007.2##
[0061] The derivative D'(x) of a function D(x) may be calculated
from
D ' ( x ) = lim z -> 0 D ( x + z ) - D ( x ) z ##EQU00008##
[0062] Estimating the vector gradient of error with respect to each
of the parameters may therefore comprise changing each parameter by
a small amount, recalculating SSE and estimating the vector
gradient of error using the difference between the reference value
for SSE and the recalculated value of SSE.
[0063] Estimating F'(t)=0 and F''(t)=0 may therefore comprise
determining two new parameter sets, the first parameter set being
changed proportionally to the estimated vector gradient of error by
addition and the second parameter set being changed proportionally
to the estimated vector gradient of error by subtraction,
calculating SSE values for each of the two new parameter sets and
calculating estimates for F'(t)=0 and F''(t)=0 from the new SSE
values.
[0064] The desired touch sensitivity may be a maximum at a first
test point and a minimum at a second test point. Alternatively, the
desired touch sensitivity may be a response which is between the
minimum or maximum at a given test position, for example, where the
responses at multiple test positions are to be taken into
account.
[0065] The desired touch sensitivity may provide the sensation of a
button click to a user. Alternatively, a touch sensitivity (in
terms of produced displacement and/or acceleration of the touch
sensitive member) may be generated to provide additional
information to the user. The filtered output signal may be
associated with a user action or gesture etc. Alternatively, or
additionally, the filtered output may be associated with the
response of the touch-sensitive surface in terms of display action
or reaction.
[0066] The vibration may include any type of vibration, including
bending wave vibration, more specifically resonant bending wave
vibration.
[0067] The transducer may be an electromagnetic transducer. Such
transducer are well known in the art. Alternatively, the exciter
may be a piezoelectric transducer or a bender or torsional
transducer. A plurality of transducer (perhaps of different types)
may be selected to operate in a co-ordinated fashion.
[0068] The touch surface may be a panel-form member which is a
bending wave device, for example, a resonant bending wave device.
The touch screen may also be a loudspeaker wherein a second
vibration exciter excites vibration which produces an acoustic
output. For example, the touch screen may be a resonant bending
wave mode loudspeaker as described in International Patent
Application WO97/09842 which is incorporated by reference.
[0069] Contact on the surface may be detected and/or tracked as
described in International patent applications WO 01/48684, WO
03/005292 and/or WO 04/053781 to the present applicant. These
International patent applications are here incorporated by
reference. Alternatively, other known methods may be used to
receive and record or sense such contacts.
[0070] According to a seventh aspect of the invention, there is
provided a touch sensitive device comprising [0071] a
touch-sensitive member, [0072] a plurality of transducers mounted
to the member, and [0073] a processor configured to carry out the
method of any one of the preceding aspects.
[0074] According to a eigth aspect of the invention, there is
provided a computer program comprising program code which, when
executed on a processor of a touch sensitive device, will cause the
touch sensitive device to carry out the method of any one of the
first to sixth aspects.
[0075] The invention further provides processor control code to
implement the above-described methods, in particular on a data
carrier such as a disk, CD- or DVD-ROM, programmed memory such as
read-only memory (firmware), or on a data carrier such as an
optical or electrical signal carrier. Code (and/or data) to
implement embodiments of the invention may comprise source, object
or executable code in a conventional programming language
(interpreted or compiled) such as C, or assembly code, code for
setting up or controlling an ASIC (Application Specific Integrated
Circuit) or FPGA (Field Programmable Gate Array), or code for a
hardware description language such as Verilog (Trade Mark) or VHDL
(Very High speed integrated circuit Hardware Description Language).
As the skilled person will appreciate such code and/or data may be
distributed between a plurality of coupled components in
communication with one another.
BRIEF DESCRIPTION OF DRAWINGS
[0076] The invention is diagrammatically illustrated, by way of
example, in the accompanying drawings in which:--
[0077] FIG. 1a is a flowchart showing the steps for improving an
estimate for a filter;
[0078] FIG. 1b is a flowchart showing the steps for an alternative
method of improving an estimate for a filter;
[0079] FIG. 2a is a graph showing the initial filter set varying
against time;
[0080] FIG. 2b is a graph of time variation of the filter set of
FIG. 2a after iteration according to the method shown in FIG.
1;
[0081] FIG. 2c is a graph of signal level (dB) against time for the
filtered response of FIG. 2a;
[0082] FIG. 2d is a graph of signal level (dB) against time for the
filtered response of FIG. 2a;
[0083] FIG. 3a is a graph of signal level (dB) against frequency
for a second initial filter set;
[0084] FIG. 3b is a graph of signal level (dB) against frequency
for the filter set of FIG. 3a after iteration according to the
method shown in FIG. 1;
[0085] FIG. 4 is a flowchart showing the steps for creating an
initial filter set;
[0086] FIG. 5a shows a log-log plot of |force| vs frequency for
each channel F and the arithmetic sum FA;
[0087] FIG. 5b shows a log-lin plot of each channel of FIG. 5a
divided by FA (i.e. normalised);
[0088] FIGS. 6a to 6d show the impulse response as it varies with
time for each of the signal of FIG. 5a;
[0089] FIG. 7a shows the time reversal filters for each of FIGS. 6a
to 6d;
[0090] FIG. 7b shows the time reversal filters of FIG. 7a convolved
with each respective signal of FIGS. 6a to 6d;
[0091] FIG. 8 is a schematic illustration of a touch sensitive
device;
[0092] FIG. 9a is a flow chart of an alternative method for
creating an initial filter set;
[0093] FIG. 9b is a graph plotting the imaginary part against the
real part for p.sub.k and -sin(.theta..sub.k) against
cos(.theta..sub.k);
[0094] FIGS. 9c and 9d plot the eight variations of u with
time;
[0095] FIG. 9e shows the variation with time of the transfer
functions for each of the four exciters, and
[0096] FIGS. 10a and 10b plot the variation in time of the quiet
and loud signals.
DETAILED DESCRIPTION OF DRAWINGS
[0097] FIG. 1. shows a flow diagram of a first example of a method
used to create an improved filter. This method involves minimising
a Summed Squared Error (SSE) value, with the SSE being used as in
example only. FIG. 1. Shows use of the gradient Search method for
the minimisation of SSE, but this is used as an exemplar method
only.
[0098] FIG. 1 shows the steps of a first method used to create an
improved filter. The filter may be a simultaneous multi-region
filter (SMR) filter or a time-reversed filter (TR) filter created
as described in co-pending International application
PCT/GB2010/050540 (the entire contents of which are incorporated
herein by reference). The first step S100 is to create an initial
estimate of the filter to be applied to a signal output from each
transducer whereby the filtered output signal is derived from
vibration of the member in response to a touch, so that the
filtered output signals from all of the transducers provide a touch
sensitivity. This signal from a transducer will be referred to as
an output signal, although it is of course an input signal from the
point of view of the filter. The output time response for the
filter is the convolution of the output signal input to the filter
and the time impulse response of the filter.
[0099] The output signal from the transducer is considered as the
output of a system model Y(n, a, b, . . . ) where a, b are the
model parameters (there may be m parameters) and n is the number of
test points at which the real output X(n) produced in response to a
touch of known magnitude is measured. Thus, as recorded at step
S102, the model is created using the model parameters,
corresponding to the taps, i.e. the coefficients on the digital
filter. As discussed above the coefficients of the filter may
correspond to the model parameters. The system model may comprise
an inverse of the filter. The initial estimate of the filter may be
used to create the model.
[0100] The next step is to measure the effectiveness of the model
by using the sum-squared error (SSE) which is defined as the energy
of the difference between the model and the real system. This is
calculated from
SSE ( a , b , ) = n Y ( n , a , b , ) - X ( n ) 2 ##EQU00009##
[0101] In other words the difference between the model output and
the real output at each test point is calculated, each difference
is squared and the squared differences are summed together. The
real output may be determined by measurement. For example, using
contact methods such as using a stylus in contact with the touch
sensitive member to cause vibration of the member by applying a
touch force of known magnitude and/or other contact methods as
would be appreciated by a person skilled in the art.
[0102] At step S104, a value for SSE is calculated that is termed
the reference SSE value (or reference value for SSE). The reference
value for SSE is calculated making use of the initial filter
estimate. This means that the reference value for SSE may be
calculated for the output of the model by comparing the output of
the model with a measured value of output signal.
[0103] At step S106, the parameters of the model are changed in
order to obtain a modelled output that tends towards the real
output. This means that the parameters of the model are changed in
order that the calculated output signals generated in response to
the touch tends towards the measured output signals generated in
response to the touch. Accordingly, each model parameter (that may
correspond to each tap or coefficient of the filter) is adjusted by
a small amount (i.e. by less than 10%, preferably less than 1%) and
a new value for SSE is calculated.
[0104] In order to obtain a model output that tends towards the
real output the aim or requirement is to minimise the SSE value.
The SSE may be minimised by any suitable minimisation routine.
Using a Gradient Search method, minimisation of SSE is achieved by
setting the gradient to zero or substantially close to zero.
[0105] For a model containing m parameters, there are m gradient
equations. Thus at step S108, the vector gradient of error is
calculated with respect to each of the parameters using
grad ( SSE ) = ( .differential. SSE .differential. a ,
.differential. SSE .differential. b , ) T ##EQU00010##
[0106] The vector gradient represents the magnitude and direction
of the slope of the error function, which in this example is the
SSE.
[0107] At step S110, two new models are created by changing the
model parameters by a value which is proportional to the calculated
vector gradient. One model output is determined by adding the value
to each of the model parameters and calculating a model output and
the second model output is determined by subtracting this value
from each of the model parameters and calculating a model output.
Using these two model outputs, two new SSEs are calculated; one
called SSE_p (from the new model where the value is added to model
parameters) and one called SSE_m (from the new model where the
value is subtracted from model parameters).
[0108] The next step S112 is to obtain a set of parameters which
aims to reduce the value of SSE. Using a Gradient Search method,
the value of SSE may be minimised by attempting to find the value
of t that minimises F(t), where F(t) is defined as:
F ( t ) = SSE ( a + t .differential. SSE .differential. a , b + t
.differential. SSE .differential. b , ) ##EQU00011##
[0109] At the value of t where F(t) is minimised, the gradient of F
is zero, i.e. for F(t) the derivative, F'(t)=0. The Newton-Raphson
method is used to solve for t, starting with t.sub.0=0. Other
equivalent known methods may be used to solve for t. For the Newton
Raphson method a new value for t (t.sub.k+1) can be calculated from
a present or start value for t (t.sub.k) from
t k + 1 = t k - F ' ( t k ) / F '' ( t k ) , where F ' ( t ) =
.differential. F .differential. t ##EQU00012## and F '' ( t ) =
.differential. 2 F .differential. t 2 ##EQU00012.2##
[0110] Starting with t=0, a first solution is
t = F ' ( 0 ) F '' ( 0 ) ##EQU00013##
[0111] The first and second derivatives of F(t) may be found using
finite difference methods. Two-sided estimates of F'(0) and F''(0)
are used to determine t, using the standard equations:
f ( x ) x ~ f ( x + x ) + f ( x - x ) 2 x and ##EQU00014## 2 f ( x
) x 2 ~ f ( x + x ) + f ( x - x ) - 2 f ( x ) x 2
##EQU00014.2##
[0112] Once a value for t is calculated, this is used to generate a
new set of model parameters and hence a new SSE value. If this
value is less than the reference value calculated in step S104, the
new SSE value is set as the reference value (step S116) and steps
S106 to S112 are repeated. If this new value is greater than the
reference value (either the value calculated in step S104 or a
previous value of SSE calculated using steps S106 to S110), the
model parameters for the improved filter are output. If the new
value is greater than the reference value, then no further
improvement can be achieved by this methodology and an optimum
solution has been achieved by the previous set of model parameter
values.
[0113] FIG. 1b shows an alternative iterative method for
calculating an improved filter. The filter may be a simultaneous
multi-region filter (SMR) filter or a time-reversed filter (TR)
filter. The method may show computational efficiencies as only one
parameter of the model need be varied within a calculation step.
Consider the general situation where we have the multi-channel
impulse responses measured at two target locations, Q is desired to
be "insensitive" and L is desired to be "sensitive".
[0114] At step S200, we define a set of filter impulse responses,
h, which represent an estimate of the filters needed to achieve the
separation of the Q and L target sensitivities. These may be an
estimate of the simultaneous multi-region filter (SMR) filter or
time-reversed filter (TR) filter.
[0115] At step S202, we calculate a reference value to be used in
calculating an improved filter, using the sum over all channels of
the convolution product of the filter impulse response at each
channel and the response measured for that channel for a touch at
the quiet location and the sum over all channels of the convolution
product of the filter impulse response at each channel and the
response measured for that channel for a touch at at the loud
location. The reference value is termed SSE. Thus the reference
value for SSE is
SSE = chan Q chan * h chan 2 chan L chan * h chan 2
##EQU00015##
where the star signifies the convolution product, Q.sub.chan is the
measured value for the quiet location, and L.sub.chan is the
measured value for the loud location.
[0116] At step S204, we "perturb" the set of filter impulse
responses h by adjusting a tap (i.e. a digital coefficient) by a
small amount a in only one filter channel j. So
h1=h, chan.noteq.j, h1=h+.alpha..delta.(t-nT), chan=j, T=sample
period
where .delta. is the delta (sampling) function and where
.delta.(x)=1 if x=0,0 otherwise
[0117] At step S206, we calculate a new value for SSE for the
perturbed value from:
SSE ( .alpha. ) = chan Q chan * ( h chan + .alpha. .delta. chan , j
z - n .delta. ( 0 ) ) 2 chan L chan * ( h chan + .alpha. .delta.
chan , j z - n .delta. ( 0 ) ) 2 ##EQU00016##
Where z is the unit delay operator .delta..sub.i,j is the Kronecker
delta which is 1 if i=j, and 0 otherwise (i.e a discrete version of
the delta function)
[0118] At step S206, we define the gradient vector exactly as
follows;
[0119] Now as the numerator and denominator for SSE(c) are
identical in form, lets consider just one term for now.
term ( .alpha. ) = chan R chan * ( h chan + .alpha. .delta. chan ,
j z - n .delta. ( 0 ) ) 2 ##EQU00017##
Where R represents either Q or L.
[0120] Expanding this explicitly, assuming that the second order
term is vanishingly small will give:
term ( .alpha. ) = chan 1 R chan 1 * ( h chan 1 + .alpha. .delta.
chan 1 , j z - n .delta. ( 0 ) ) chan 2 R chan 2 * ( h chan 2 +
.alpha. .delta. chan 2 , j z - n .delta. ( 0 ) ) ##EQU00018## term
( .alpha. ) = term ( 0 ) + 2 .alpha. ( z - n R j ) ( chan R chan *
h chan ) + O ( .alpha. 2 ) ##EQU00018.2##
[0121] Using this result, we may define the exact gradient of SSE
as follows;
grad ( SSE ) n , j = lim .alpha. -> 0 ( SSE ( .alpha. ) - SSE
.alpha. ) ##EQU00019## grad ( SSE ) n , j = 2 chan L chan * h chan
2 ( ( z - n Q j ) ( chan Q chan * h chan ) - SSE ( z - n L j ) (
chan L chan * h chan ) ) ##EQU00019.2##
[0122] Now we have only 2 convolutions and some shifts and dot
products, and the result is exact. Thus at Step S210 we calculate
the value at which the gradient vector and follow the same final
steps as in the previously described method.
[0123] FIGS. 2a to 2d show the method of FIG. 1 applied to a
specific example, namely a TR filter (see FIGS. 4 to 8). The TR
filter produces a simultaneous minimum and maximum, the said Quiet
and Loud signals produced by touches at two separate points having
respective lower and higher relative sensitivities for a touch
sensitive device having four transducers in this specific example,
however as appreciated by the person skilled in the art the claimed
method and device may operate with other numbers of transducers,
such as 2 or more transducers. FIG. 2a shows the filter for each of
the four transducers which is essentially a time-reversed impulse
response. FIG. 2c shows the output time response of the filters of
FIG. 2a (i.e. shows the convolution of the impulse responses of the
filter with the input signals). The maximum signal is shown as
testABL.sub.k and the minimum signal as testABQ.sub.k. The SSE for
this filter set is -2.6 dB which is better than creating a filter
at random but not significantly so.
[0124] FIG. 2b shows the filter for each of the four transducers
after application of ten iterations of the optimisation method of
FIG. 1. FIG. 2d shows the output time response of the filters of
FIG. 2c. The SSE for this filter set is -17.9 dB. This translates
in to a clearly visible difference between the maximum output of
the maximum signal (testABL.sub.k) and the maximum output of the
minimum signal (testABQ.sub.k). The difference is considerably
greater than the same difference for the original filters and
renders the optimised filters effective whereas the original
filters were not very effective.
[0125] FIGS. 3a and 3b show the application of the optimisation
method of FIG. 1 to an SMR filter which has been obtained
analytically using the eigenvector method. These show two traces
for output signals produced by touches at the loud position and two
traces for output signals produced by touches at the quiet
position, both before and after application of the optimisation
method.
[0126] One set of filters is created for a net-book portable
computer with a touch-screen by dividing the touch screen into a
grid having three rows and five columns. Measurements are taken for
touches at the "2,2" target (i.e. second row, second column) and
the "3,1" target to generate two sets of SMR filters with the goal
of minimising the response at one target and maximising it at the
other, that is, minimising the sensitivity at one target and
maximising the sensitivity at the other. The filtered output
signals resulting from applying the calculated filters to an output
signal were measured and are shown in FIG. 3a. For both filter
sets, the intendedly loud signals (50, 52) are stronger than the
intendedly quiet signals (60, 62). However, at some frequencies the
difference is less than desired. Moreover, such SMR filters are
exact in the frequency domain but seldom achieve a theoretical
separation better than 20 dB in the time-domain. This can be
improved somewhat by improving the frequency resolution of the
measurement data and by adjusting the delay of the filters.
However, there still seems to be a fundamental limit to the
effectiveness of the filters in the time-domain. One explanation
may be the Heisenberg uncertainty relationship between frequency
and time means that without an infinitely long filter and an
infinitely fine frequency resolution, the time-domain filters
obtained this way will never be exact.
[0127] FIG. 3b shows the result of iteratively refining the initial
filter set according to the method of FIG. 1. In real use, the
filters still have to contend with system non-linearity and time
variation but the measured responses still show a very useful
improvement in performance. The output filtered signals resulting
from applying the optimised filters to an input signal were
measured and are shown in FIG. 3b. The separation between responses
to touches at the loud and quiet locations is more marked with the
refined filter set than with the original set. The refinement
process has effectively doubled the dB separation of the
analytically derived filters. In real use, the filters still have
to contend with system non-linearity and time variation but the
measured responses still show a very useful improvement in
performance.
[0128] Moreover, it can be shown that the process makes empirically
derived SMR filters effective.
[0129] The "exact" solutions, i.e. the analytically calculated SMR
filters, result in filter delays of about 1/10.sup.th of the delays
from the time-reversal solution and may be preferable. Time
reversal filters may intrinsically incur temporal delays, as shown
in FIG. 2a.
[0130] FIGS. 4 to 7b illustrate one method for creating an initial
filter set, in this case using a time-reversed impulse response. As
explained in more detail below, time-reversed impulse response (TR)
shows how to create a single maximum sensitivity. With a minimum
amount of effort, it can also give a single minimum sensitivity. To
get simultaneously min and max sensitivities at different locations
would probably use the minimising set, and could either rely on
there being some sensitivity at the other location, or require
extra "empirical" combining of the minimising and maximising
filters. Thus, it is possible to empirically derive an SMR using
the TR process. The method of improving the filter set described in
relation to FIGS. 1 to 3b could be used in relation to any filter
set, including conceptually a blind guess. However, the better the
initial estimate, the better the result.
[0131] As shown in FIG. 4, the first step S200 in creating a filter
set is to physically touch the screen with a known force at test
positions and to measure the resulting output signal at this
plurality of locations (S202). As explained with reference to FIG.
5a, each measured response is optionally whitened (S204) and then
transformed into the time domain (S206). As explained in FIGS. 6a
to 6d, the filter is formed by taking a snapshot of each impulse
response (S208) and reversing this snapshot (S210).
[0132] The spectrum of the time-reversed signal is the complex
conjugate of the original
x(t)->X(f) original
y(t)=x(-t); Y(f)=conj(X(f)) filter
[0133] This is approximated by adding a fixed delay, so
z(t)=x(T-t) if t<=T, or z(t)=0 if t>T
[0134] When the filter is applied to the signal (ignoring the
approximation for now), the phase information is removed, but the
amplitude information is reinforced.
y(t)*x(t)->X(f).times.Y(f)=|X(f)| 2
[0135] (In fact, the resulting time response is the autocorrelation
function).
[0136] FIG. 5a is a log-log plot of power vs frequency for each
channel (F) (i.e. for signals from four vibration transducers) and
the arithmetic sum (FA) of these four power signals. FIG. 5b is a
log-lin plot of each of the 4 channel responses divided by FA (FN),
i.e. normalised. Dividing each response by FA renders them more
spectrally white, which improves the effectiveness of the method.
This is because the response is squared in the filtering process
and thus it is beneficial if the signal is spectrally "white", or
flat.
[0137] FIGS. 6a to 6d show each of the four normalised responses
transformed into the time domain to give the individual impulse
responses (GYtime.sup.<1>). These impulse responses have also
been normalised by dividing by the peak for each response.
[0138] The time reversal filters, TR, are formed by taking a finite
snapshot of the impulse responses of FIGS. 6a to 6d and then
reversing them in time. The TR have built into them a delay equal
to the length of the sample, as shown in FIG. 2a. FIG. 7a shows the
time reveral filter for each channel (0, 1, 2 and 3)
TR k , j := if ( k .gtoreq. k max , 0 , GYtime kmax - k - 1 , j + 1
win kmax - k - 1 ) peak j + 1 ##EQU00020##
where kmax=samples (length)
[0139] FIG. 7b shows the results of convolved the filters with the
appropriate impulse responses in the time domain (GYtime) to give
the filtered response GYresp. The convolution is expressed as:
GYresp k , j := i = 0 k ( GYtime k - i , j + 1 TR i , j )
##EQU00021## amp j := max ( GYresp j ) ##EQU00021.2##
[0140] As shown in FIG. 7b, all the responses share a common
maximum, but exhibit some ringing. The common maximum occurs
because the phase/time information has been corrected. The ringing
occurs because the amplitude information is exaggerated.
[0141] As shown at step S212 in FIG. 4, the filter amplitudes may
be adjusted to maximise or minimise the sum of the four signals
produced from the touch location. The filter is then applied to
each impulse response to generate a filtered output signal derived
from each location (S214).
[0142] FIG. 8 shows a touch-sensitive haptics device 10 with four
haptics-input and touch sensing vibration transducers 12 mounted to
a touch-sensitive screen 14 (in this specific example there are
four transducers, however as appreciated by the person skilled in
the art there may be any number of input transducers on the
screen). The transducers 12 are each coupled to a system processor
20 via a two-way amplifier 22. A stylus 16 is also connected to the
processor 20 via a two-way amplifier 24.
[0143] The touch-sensitive device shown in FIG. 8 may be used to
create an initial filter set as set out in FIG. 4 and may then be
used to apply the improved filter set created as described in FIG.
1. The device has two operational modes, normal use and training
mode. In normal use, i.e. when a user is using the screen 14 of the
touch-sensitive device 10, the transducers 12 produce output
signals in response to touches on the surface. These output signals
can then be used to determine properties of the touches such as
location and force of touch. The method of determining these
properties is not critical to the operation of the device and may
be as described in any known techniques. The output signals can be
processed to control touch sensitivity of the device. The touch
sensitivity may be simple, for example detecting touches only in
specific locations, or may be more complex to identify complex
touch activity, i.e. associated with sliding movements,
increasing/decreasing intensity of touch etc. The more complex
sensations may be associated with gestures such as sliding,
pitching or rotating fingers on the screen.
[0144] The transducers 12 also produce any required localized
haptic force feedback.
[0145] The transducers are thus reciprocal transducers able to work
as both output devices to generate excitation signals which create
vibration in the screen and as input devices to sense vibration in
the screen and convert the vibration into signals to be analysed.
It is preferable for all the transducer to be reciprocal devices
but it is possible to have a device in which not all transducers
are reciprocal; such a device is more complicated.
[0146] In training mode, the stylus 16 is used to inject
vibrational signals at specified test points; thus the stylus 16
may be considered to be a "force pencil". The system processor 20
generates the signals which are sent to the stylus 16 via the
two-way amplifier 24 and receives the signals from the transducers
12. The two-way amplifiers 22 are also connected between the system
processor 20 and each transducer 12; one amplifier for each
channel, i.e. one amplifier for each transducer. The stylus 16 is
also arranged to sense haptic feedback signals in the screen
originating from the transducers 12 and to feed the sensed signals
to the processor 20 via the two-way amplifier 24.
[0147] FIGS. 9a to 10b illustrate an alternative method for
creating an initial filter set, in this case a simultaneous
multi-region filter (SMR) having a maximum touch sensitivity at one
location on the touch sensitive screen and a minimum touch
sensitivity at another discrete location on the screen. Four
transducers are used in the filter so there are four channels with
j:=0 . . . chan-1. In this specific example there are four
transducers, however as appreciated by the person skilled in the
art the claimed method and device may operate with 2 or more
transducers.
[0148] The alternative method for creating an initial filter set
may be carried in two ways.
[0149] A first way may be used when the filter impulse response is
known. This means that the required theoretical filter impulse
response to give the desired member touch sensitivity output may be
known, It will be understood that even if the desired impulse
response is known, the implementation of the required filter in
practice may be difficult, as will be appreciated by the person
skilled in the art. The solution may be to make use of Quiet and
Loud measurements made by applying a touch to different points on
the member or panel, and a generalised inverse matrix technique
(also known as the Moore-Penrose technique) may be used to
determine recursive filters, the output from which may be used to
provide a filtered impulse response tending towards the required
theoretical impulse response to give the desired touch sensitivity
response.
[0150] The second way may be used even when the filter impulse
response is not known. The solution may be to use to make use of
Quiet and Loud measurements made by applying a touch to different
points on the member or panel, and a solving an eigenvalue problem
may be used to determine recursive filters, the output from which
may be used to provide a filtered impulse response tending towards
the required theoretical impulse response to give the desired
vibration output.
[0151] The two ways of creating an initial filter set using an
alternative method have a number of common steps, before the
generalised inverse matrix technique or solution of an eigenvalue
problem are undertaken.
[0152] It should be understood that although the terms quiet and
loud are used to refer to locations having relatively less and more
sensitivity to a touch, producing a desired touch sensitivity
profile, this wording is used only to assist understanding. It is
not necessary that the touch produces an audible sound.
[0153] As set out in FIG. 9a, the first step S300 is to choose a
set of frequencies spanning the frequency range of interest. The
set of frequencies may be linearly, logarithmically or otherwise
nonlinearly distributed across the frequency range of interest. For
example, the frequency range of the filter may be from 150 Hz (bot)
to 600 Hz (top). Q (and q) is an abbreviation for quiet (i.e. the
minimum touch sensitivity and response) and L (and 1) is an
abbreviation for loud (i.e. the maximum touch sensitivity and
response). Thus the initial variables are defined as:
Time = Qdata 0 ##EQU00022## chans = col ( Qdata ) - 1
##EQU00022.2## q j = Qdata j + 1 ##EQU00022.3## l j = Ldata j + 1
##EQU00022.4## F s = length ( time ) - 1 time last ( time ) - time
0 ##EQU00022.5## F s = 5512.5 ( the sampling frequency )
##EQU00022.6##
[0154] An IIR filter is used to create the desired effect. An IIR
is an infinite impulse response filter. Such filters use feedback
since the output and next internal state are determined from a
linear combination of the previous inputs and outputs. A second
order IIR filter is often termed a biquad because its transfer
function is the ratio of two quadratic functions, i.e.
H z ( z , d ) = k d k , 0 + d k , 1 z - 1 + d k , 2 z - 2 ( 1 - p k
z - 1 ) ( 1 - p _ k z - 1 ) ##EQU00023##
[0155] The transfer function of such a filter has two poles and two
zeros. A pole of a function f(z) is a point a such that f(z)
approaches infinity as z approaches a and a zero is a point b such
that f(z) equals zero when z equals b. Thus, the d.sub.k,0
d.sub.k,1 d.sub.k,2 co-efficients determine the zeros and the
p.sub.k coefficients determines the poles. The set of frequencies
within the frequency range of interest is associated with a
corresponding pole, where k may equal the number of frequencies
within the set, equal to the number of poles.
[0156] The pole coefficients p.sub.k may be written as:
p.sub.k:=e.sup.-0.5.DELTA..theta..sup.ke.sup.-j.theta..sup.k
[0157] As shown in FIG. 9a, the next step S302 of the method is to
calculate .theta..sub.k and .DELTA..theta..sub.k which may be
derived from the initial variables as follows:
K := floor ( per log ( top bot , 2 ) + 0.5 ) ##EQU00024## K = 8
##EQU00024.2## k := 0 K - 1 ##EQU00024.3## k 1 := 1 K - 2
##EQU00024.4## f k := bot ( top bot ) k K - 1 ##EQU00024.5## f K -
1 := top ##EQU00024.6## .theta. k := 2 .pi. f k Fs ##EQU00024.7##
.DELTA. .theta. K - 1 := .theta. K - 1 - .theta. K - 2
##EQU00024.8## .DELTA. .theta. 0 := .theta. 1 - .theta. 0
##EQU00024.9## .DELTA. .theta. k 1 := .theta. k 1 + 1 - .theta. k 1
- 1 2 ##EQU00024.10##
Where K is the number of poles, f.sub.k is the frequency of the
k.sup.th pole, F.sub.s is the sampling frequency, .theta..sub.k is
an angle related to the frequency of the k.sup.th pole
[0158] FIG. 9b plots the value of the imaginary part of the pole
coefficient p.sub.k against the real part of p.sub.k. FIG. 9b also
plots the value of -sin(.theta..sub.k) against cos(.theta..sub.k)
and shows the locus of pole positions p.sub.k plotted along with an
arc of the unit circle in the z-plane.
[0159] The p.sub.k values are complex. If we wish to consider real
coefficient values, the transfer function may be written as:
H z ( z , d ) := k ( d k , 0 + d k , 1 z - 1 + d k , 2 z - 2 1 + a
k , 0 z - 1 + a k , 1 z - 2 ) ##EQU00025##
[0160] Now the a.sub.k,0 a.sub.k,1 coefficients determine the
poles. These coefficients may be written:
a k , 0 := - 2 - .DELTA. .theta. k 2 cos ( .theta. k ) ##EQU00026##
a k , 1 := - .DELTA. .theta. k ##EQU00026.2## i . e . a k , 0 = - 2
Re ( p k ) and a k , 1 = p k 2 ##EQU00026.3##
[0161] At this step we also solve for the poles by considering:
Y X = 1 1 + a 0 z - 1 + a 1 z - 2 ##EQU00027## Y ( 1 + a 0 z - 1 +
a 1 z - 2 ) = X ##EQU00027.2## Y = X - ( a 0 z - 1 + a 1 z - 2 ) Y
##EQU00027.3##
[0162] For input X(z), the result of filtering with only the poles
is Y(z), hence the transfer function is Y(z)/X(z). This is then
algebraically manipulated to get a result without division, which
can be directly converted to the time-domain representation u(i,k)
which are defined as follows:
u.sub.0,k:=1 u.sub.1,k:=-(a.sub.k,0u.sub.0,k)
u.sub.i+2,k:=-(a.sub.k,0u.sub.i+1,k+a.sub.k,1u.sub.i,k)
[0163] The variations in these functions with time are plotted in
FIGS. 9c and 9d. The next step is to solve for the zeros, i.e. to
determine the coefficients d.sub.k,0 d.sub.k,1 etc. The zeros may
be paired with the poles or they may be a different set of zeros.
The zeros are determined from:
- 1 2 d k , 0 [ d k , 1 - ( d k , 1 ) 2 - 4 d k , 0 d k , 2 d k , 1
+ ( d k , 1 ) 2 - 4 d k , 0 d k , 2 ] ##EQU00028##
[0164] If the impulse response X(z) is known as shown in FIG. 9a at
step S304, the unique solution may be found at step S306, e.g. by
using the generalised inverse matrix which is defined as
follows:
M.sub.i+.DELTA.,3k:=u.sub.i,k M.sub.i+.DELTA.+1,3k+1:=u.sub.i,k
M.sub.i+.DELTA.+2,3k+2:=u.sub.i,k M.sub.i,3K+m:=.delta.(i,m)
M:=submatrix(M,0,last(time,0,cols(M)-1) MI:=geninv(M)
[0165] The zero coefficients are then
d.sup.<j>:=MIh.sup.<j>
[0166] The calculated matrix is shown below:
d = 0 1 2 3 0 3.405 2.672 42.262 15.453 1 - 0.537 - 3.023 4.175
1.281 2 - 2.112 - 2.489 - 35.365 - 18.147 3 2.287 7.815 60.95 -
0.504 4 - 0.437 - 3.615 7.214 - 0.873 5 - 0.743 - 8.985 - 48.999 -
5.914 6 20.899 3.204 49.812 - 12.642 7 - 0.587 - 2.456 13.298 -
1.674 8 - 20.614 - 1.808 - 25.871 5.237 9 10.879 10.881 - 6.574 -
40.342 10 - 1.759 - 6.355 11.212 - 5.183 11 - 12.464 - 17.547
29.148 27.938 12 - 24.895 12.807 - 4.562 - 75.415 13 - 12.819 -
5.264 - 15.278 0.249 14 4.693 - 17.369 - 24.445 76.935 15 - 9.55
6.87 32.191 ##EQU00029##
Rows(p)=8 with Max(p)=09.67-0.167j and min(p)=0.729-0.594j
[0167] The pairs of zeros associated with the corresponding d
values are thus:
z 1 k , j := d 3 k + 1 , j - ( d 3 k + 1 , j ) 2 - 4 d 3 k , j d 3
k + 2 , j - 2 d 3 k , j ##EQU00030## z 2 k , j := d 3 k + 1 , j + (
d 3 k + 1 , j ) 2 - 4 d 3 k , j d 3 k + 2 , j - 2 d 3 k , j
##EQU00030.2##
[0168] The pairs of zeros relates to the zero coefficients d in the
same way that the poles p relate to the pole coefficients a except
that the zeros cannot be asserted to appear as complex conjugate
pairs.
[0169] The synthesised transfer function (hs) for each filter to be
applied to the output signal from each transducer to give the
required sensitivity profile to touches of the member may then be
determined from:
hs .DELTA. + i , j := k ( uf ( i , k ) d 3 k , j + uf ( i - 1 , k )
d 3 k + 1 , j + uf ( 1 - 2 , k ) d 3 k + 2 , j ) ##EQU00031##
[0170] As set out above, there may be zeros which are not paired
with poles. In this case, the transfer function may be written
as:
H z ( z , d , b ) := z - .DELTA. k ( d k , 0 + d k , 1 z - 1 + d k
, 2 z - 2 1 + a k , 0 z - 1 + a k , 1 z - 2 ) + m ( b m z - m )
##EQU00032##
[0171] Where the b terms are coefficients which define additional
zeros with no associated poles. Such an additional term may be
implemented as an additional filter which may typically be used to
deal with excess-phase problems such as delay. In other words, each
filter may comprise at least two separate filters which may be of
different types. The individual transfer functions for each filter
may be expressed as:
hs i , j := hs i , j + m ( d 3 K + m , j .delta. ( i , m ) )
##EQU00033##
[0172] Where the first term for hs sums all the contributions from
the first type of filter (e.g. biquad) and the second term adds in
the additional contributions from a second filter (e.g. a
discrete-time filter such as a finite impulse response (FIR)).
[0173] Alternatively, if the impulse response is not known, a
solution may be derived as follows.
[0174] Consider Quiet and Loud impulse responses to be determined
for touches at two locations on the member. If the Quiet response
is minimal or zero (this means no response to touching the member
at the Quiet location) and the Loud response is the loudest
possible response (this means the largest possible response to
touching of the member at the Loud location), then there are a
large number of possible impulse response filters for each
transducer channel that can be used to obtain the Quiet response,
whilst there is only one set of impulse response filters that gives
the loudest possible response. Each set of filters that gives the
minimal or zero quiet response to a touch at the quiet location
will also give some response (which may be minimal or zero or may
be non-zero) to a touch at the loud location. However, the set of
filters yielding the loudest possible response to a touch at the
loud position may not, and probably will not, yield minimal or zero
output to a touch at the quiet location.
[0175] There is a non-unique set of solutions relating to sets of
filters that yields a minimal or zero quite response to a touch at
the quiet location, whilst responding to touching of the member at
the loud location. There may be no solution relating to sets of
filters that yields a minimal or zero quiet response to a touch at
the quiet location whilst yielding a maximum response signal to a
touch at the loud location.
[0176] The problem of finding an improved set of filters relates to
finding the set of filters that maximises the loud signal in
response to a touch at the loud location and at the same time
minimises the quiet signal in response to a touch at the quiet
location. The problem may be solved by considering the ratio
between the quiet and loud response signals, and choosing a set of
filters that yields a minimal or zero response at the quiet
location, determine the response at the quiet location, determine
the associated response at the loud location for the same set of
filters, and determine the ratio between the signals at the quiet
and loud locations. By considering one, all or any of the sets of
filters that yields a minimal or zero response at the quiet
location the set of filters that minimises the ratio between the
quiet response at the quiet location and the loud response at the
loud location may be selected. These sets of filters may be
considered as the improved set of filters.
[0177] Further, linear combinations of the sets of filters that
individually yield a minimal or zero response to a touch at the
quiet location will also yield a minimal or zero response to a
touch at the quiet location. Accordingly, the problem may be solved
by selecting a linear combination of different ones of the sets of
filters that yield a minimal or zero response to a touch at the
quiet location, the linear combination being selected to maximise
the response to a touch at the loud location.
[0178] Mathematically, a non-unique solution may be derived, for
example by solving the eigenvalue problem as set out below at Step
S308. Where we are deriving the non-unique solution, additional
requirements may be imposed and only the combination of solutions
that satisfies the additional requirements is generated (step
S310).
[0179] Firstly we consider the quiet response:
MQ := | M length ( time ) + 1 , 3 K chans - 1 .rarw. 0 for k
.di-elect cons. 0 K - 1 for j .di-elect cons. 0 chans - 1 for n
.di-elect cons. 0 2 for k ' .di-elect cons. 0 K - 1 for j '
.di-elect cons. 0 chans - 1 for n ' .di-elect cons. 0 2 MQ ( j ' K
+ k ' ) 3 + n ' , ( j K + k ) 3 + n .rarw. i = 0 last ( qs k , j )
( sub ( qs k , j , i - n ) sub ( qs k ' , j ' , i - n ' ) ) MQ
##EQU00034##
Where:
[0180] qs.sub.k,j:=convol(u.sup.<k>,q.sup.<j>) and
sub(v,n):=if(n<0,0,v.sub.n)
Where q.sup.<j> are the determined quiet responses for the j
channels. Simultaneously we consider the loud response:
ML.sub.n',n:=hls.sub.nhls.sub.n'
Where
[0181] hls n := j convol [ ( hs n ) j l j ] ##EQU00035##
With
[0182] hs n := | g .rarw. E n + pick for i .di-elect cons. 0 rows (
u ) - 1 for j .di-elect cons. 0 chans - 1 hsn i , j .rarw. k [ r =
0 2 [ g ( j K k ) 3 r sub ( u k i - 1 ) ] ] hsn ##EQU00036##
and
E := Re ( eigenvecs ( MQ ) ) ##EQU00037## .lamda. := eigenvals ( MQ
) ##EQU00037.2## pick := K chans 3 2 - 1 ##EQU00037.3## n := 0 K
chans 3 - pick - 1 ##EQU00037.4## n ' := 0 K chans 3 - pick - 1
##EQU00037.5## EL := eigenvecs ( ML ) ##EQU00037.6## .lamda. L :=
eigenvals ( ML ) ##EQU00037.7## c := EL 0 ##EQU00037.8## d := n ( c
n E n + pick ) ##EQU00037.9##
[0183] The transfer function for each transducer is plotted in FIG.
9e and is calculated from:
hs i , j := k [ n = 0 2 [ d ( j K + k ) 3 + n sub ( u k i - n ) ] ]
##EQU00038##
[0184] FIG. 10a plots the quiet and loud responses against time
which are produced using these transfer functions. They are
calculated from:
Q := j convol ( hs j , q j ) ##EQU00039## L := j convol ( hs j , l
j ) ##EQU00039.2## Q L = 3.887 .times. 10 - 3 ##EQU00039.3##
[0185] The maximum response has a magnitude which is approximate
4000 times greater than the minimum response.
[0186] FIG. 10b shows the results showing the filtered output
signal from each transducer.
[0187] Again the quiet and loud responses are plotted against time
and they are calculated from:
Q := convol ( TB , Q ) ##EQU00040## L := convol ( TB , L )
##EQU00040.2## Q L = 2.382 .times. 10 - 3 ##EQU00040.3##
[0188] The maximum response has a magnitude which is approximate
2000 times greater than the minimum response.
[0189] The apparatus described above may be implemented at least in
part in software. Those skilled in the art will appreciate that the
apparatus described above may be implemented using general purpose
computer equipment or using bespoke equipment.
[0190] The hardware elements, operating systems and programming
languages of such computers are conventional in nature, and it is
presumed that those skilled in the art are adequately familiar
therewith. Of course, the computing functions may be implemented in
a distributed fashion on a number of similar platforms, to
distribute the processing load.
[0191] Here, aspects of the methods and apparatuses described
herein can be executed on a mobile device and on a computing device
such as a server. Program aspects of the technology can be thought
of as "products" or "articles of manufacture" typically in the form
of executable code and/or associated data that is carried on or
embodied in a type of machine readable medium. "Storage" type media
include any or all of the memory of the mobile stations, computers,
processors or the like, or associated modules thereof, such as
various semiconductor memories, tape drives, disk drives, and the
like, which may provide storage at any time for the software
programming. All or portions of the software may at times be
communicated through the Internet or various other
telecommunications networks. Such communications, for example, may
enable loading of the software from one computer or processor into
another computer or processor. Thus, another type of media that may
bear the software elements includes optical, electrical and
electromagnetic waves, such as used across physical interfaces
between local devices, through wired and optical landline networks
and over various air-links. The physical elements that carry such
waves, such as wired or wireless links, optical links or the like,
also may be considered as media bearing the software. As used
herein, unless restricted to tangible non-transitory "storage"
media, terms such as computer or machine "readable medium" refer to
any medium that participates in providing instructions to a
processor for execution.
[0192] Hence, a machine readable medium may take many forms,
including but not limited to, a tangible storage carrier, a carrier
wave medium or physical transaction medium. Non-volatile storage
media include, for example, optical or magnetic disks, such as any
of the storage devices in computer(s) or the like, such as may be
used to implement the encoder, the decoder, etc. shown in the
drawings. Volatile storage media include dynamic memory, such as
the main memory of a computer platform. Tangible transmission media
include coaxial cables; copper wire and fiber optics, including the
wires that comprise the bus within a computer system. Carrier-wave
transmission media can take the form of electric or electromagnetic
signals, or acoustic or light waves such as those generated during
radio frequency (RF) and infrared (IR) data communications. Common
forms of computer-readable media therefore include for example: a
floppy disk, a flexible disk, hard disk, magnetic tape, any other
magnetic medium, a CD-ROM, DVD or DVD-ROM, any other optical
medium, punch cards, paper tape, any other physical storage medium
with patterns of holes, a RAM, a PROM and EPROM, a FLASH-EPROM, any
other memory chip or cartridge, a carrier wave transporting data or
instructions, cables or links transporting such a carrier wave, or
any other medium from which a computer can read programming code
and/or data. Many of these forms of computer readable media may be
involved in carrying one or more sequences of one or more
instructions to a processor for execution.
[0193] Those skilled in the art will appreciate that while the
foregoing has described what are considered to be the best mode
and, where appropriate, other modes of performing the invention,
the invention should not be limited to specific apparatus
configurations or method steps disclosed in this description of the
preferred embodiment. It is understood that various modifications
may be made therein and that the subject matter disclosed herein
may be implemented in various forms and examples, and that the
teachings may be applied in numerous applications, only some of
which have been described herein. It is intended by the following
claims to claim any and all applications, modifications and
variations that fall within the true scope of the present
teachings. Those skilled in the art will recognize that the
invention has a broad range of applications, and that the
embodiments may take a wide range of modifications without
departing from the inventive concept as defined in the appended
claims.
[0194] Although the present invention has been described in terms
of specific exemplary embodiments, it will be appreciated that
various modifications, alterations and/or combinations of features
disclosed herein will be apparent to those skilled in the art
without departing from the spirit and scope of the invention as set
forth in the following claims.
* * * * *