Patent Grant
9319804
The present invention is related to a method for operating a hearing device as well as to a hearing device.
Various approaches for frequency lowering have been pursued in order that hearing impaired patients with high frequency hearing loss can benefit, especially in those cases where the amplification of the original high frequency sound is not useful—e.g. due to dead regions—or not possible—due to potential feedback problems on applying high gain or due to limited bandwidth of applied gain.
Known teachings describing frequency lowering schemes are disclosed in WO 2007/000161 A1, U.S. Pat. No. 7,248,711 B2 and in EP-1 686 566 A2, for example. The known teachings have one or several of the following disadvantages:
It is therefore one object of the present invention to overcome at least one of the above-mentioned disadvantages.
In the context of the present invention, the term “transposition” or “transpose” is defined as having at least one of the following meanings:
Furthermore, the term “hearing device” is not only directed to hearing aids that are used to improve the hearing of hearing impaired patients but also to any communication device, be it wired or wireless, or to hearing protection device, wherein hearing aids may also be implantable.
The present invention is first directed to a method for operating a hearing device by applying a frequency transposition scheme to an input signal of the hearing device. The hearing device comprises an input transducer, a signal processing unit and an output transducer. The method according to the present invention comprises the steps of:
In a more specific embodiment of the present invention, the momentary characteristic is at least one of the following:
In further embodiments of the present invention, the source region comprises a lower source region and at least two source stacks, the lower source region being below a cut-off frequency and the at least two source stacks being above the cut-off frequency, and wherein the destination region comprises a lower destination region and a destination stack, the lower destination region being below the cut-off frequency and the destination stack being above the cut-off frequency, the cut-off frequency particularly being below 1′500 Hz.
In further embodiments of the present invention, the step of transposing comprises the following steps:
In still further embodiments of the present invention, the source region above the cut-off frequency is divided into equally sized source stacks, each having a frequency range that is equal to a frequency range of the destination stack.
In further embodiments of the present invention, the step of transposing comprises one of the following steps:
Further embodiments of the present invention comprise the step of applying a pre-weighting function to signal components of the source region before the step of adaptively selecting signal components of the source region.
In further embodiments of the present invention, the pre-weighting function is based on at least one of the following criterions:
Further embodiments of the present invention comprise the step of applying a post-weighting function to the destination region after the step of transposing the selected signal components.
In further embodiments of the present invention, the steps of selecting and transposing comprise a peak picking according to the following scheme:
i=arg max [w(n)·Fin(n),w(n+j)·Fin(n+j),w(n+j+1)·Fin(n+j+1), . . . ,w(n+j+k)·Fin(n+j+k)]
Fout(n)=w″(i)·Fin(i)
wherein
In further embodiments of the present invention, the frequency transposition scheme is defined by the following formula:
wherein
Furthermore, the present invention is also directed to a hearing device comprising:
In a more specific embodiment of the present invention, the momentary characteristic is at least one of the following:
In further embodiments of the present invention, the source region comprises a lower source region and at least two source stacks, the lower source region being below a cut-off frequency and the at least two source stacks being above the cut-off frequency, and wherein the destination region comprises a lower destination region and a destination stack, the lower destination region being below the cut-off frequency and the destination stack being above the cut-off frequency, the cut-off frequency particularly being below 1′500 Hz.
In still further embodiments of the present invention, the means for transposing comprise the following:
In further embodiments of the present invention, the source region above the cut-off frequency is divided into equally sized source stacks, each having a frequency range that is equal to a frequency range of the destination stack.
In further embodiments of the present invention, the means for transposing comprise one of the following:
Further embodiments of the present invention comprise means for applying a pre-weighting function to signal components of the source region before adaptively selecting signal components of the source region.
Further embodiments of the present invention comprise means for applying a post-weighting function to the destination region after transposing the selected signal components.
It is expressly pointed out that the above-mentioned embodiments can be arbitrarily combined. Only those combinations are excluded that would result in a contradiction.
The present invention is further illustrated by way of exemplified embodiments shown in drawings and described in detail. It is pointed out that these embodiments are for illustrative purposes only and shall not confine the present invention.
In
In case a signal processing algorithm, which is implemented in the signal processing unit 3, is applied in the frequency domain, a transformation function, such as a Fast Fourier Transformation (FFT), is used to transform the input signal i from the time domain into the frequency domain. Consequently, an inverse transformation function must be applied in order to transform an output spectrum into the time domain after implementing the signal processing algorithm. Instead of a Fourier transformation function and its inverse function, any other transformation function may be implemented, such as a Hadamard, a Paley or Slant transformation.
In particular, the present invention is directed to a signal processing algorithm (also called frequency transposition scheme) that is implemented in the signal processing unit 3. By the frequency transposition scheme, selected frequency ranges, which are important for a user of the hearing device but in which frequency ranges the user is not able to perceive an acoustic signal due to a complete hearing loss, for example, are transposed to another frequency range in which the hearing device user can perceive an acoustic signal.
Furthermore, known frequency transposition algorithms distort the harmonic structure of the input sound. Therefore, it is also not very useful for transposing music where it introduces unpleasant pitch distortions.
In connection with known frequency transposition schemes, it has been pointed out that the cut-off frequency FC must be equal or larger than 1′500 Hz in order not to distort vowels and non-fricative sounds which have a strong formant structure in a frequency region below 1′500 Hz. Therefore, signal components below the cut-off frequency FC are not changed, i.e. a so called lower source region 10 on the x-axis is linearly transposed to a lower target region 12 on the y-axis (one-to-one mapping). Above the cut-off frequency FC, a non-linear transposition is implemented in that—provided a logarithmic scale is used on the x—as well as on the y-axis—signal components of a so called higher source region 11 are transposed to a higher target region 13 that has a smaller bandwidth than the higher source region 11. The known technique does not enable a hearing impaired person to benefit from a frequency lowering algorithm having a cut-off frequency FC below 1′500 Hz, while offering acceptable sound quality and minimal distortion of vowels and non transposed sounds, which are otherwise audible without much distortion.
The present invention comprises a new frequency transposition scheme by adaptively selecting signal components of a source region taking into account momentary characteristics of the input signal.
A momentary characteristic of the input signal may be one or several of the following characteristic:
With regard to the presence of a perturbation signal, it is pointed out that if a perturbation signal is detected, the frequency range, in which the perturbation signal is present, can be excluded from being transposed to a destination region. Therewith, the destination region will contain less disturbing signal components for the hearing device user.
In further embodiments of the present invention, a so called frequency stacking algorithm is implemented.
A destination region 30 on the y-axis comprises a lower destination region 31 and a destination stack 32, the lower destination region 31 comprising frequencies up to the cut-off frequency FC, and the destination stack 32 comprising frequencies above the cut-off frequency FC.
As can be seen from
If a frequency range of a source region being transposed is equal to a frequency range of a destination region, this is called linear frequency transposition. If, on the other hand, a frequency range of a source region being transposed is greater than a frequency range of a destination region, this is called compressive frequency transposition.
As a result of these definitions,
In contrast thereto,
It is pointed out that the inventive frequency transposition scheme is very flexible in that source stacks 22 and 23 may be of any size, in particular the source stack 23 may have a larger frequency range than the one of the source stack 22.
As more than one source stacks 22 to 26 are transposed to the same destination stack 32, a yet to be described frequency transposition scheme must be applied in order to obtain a good result in the output frequency range fout, i.e. in the destination region 30, for the hearing device user.
In a specific embodiment of the present invention, the source stacks 22 to 26 have the same size. In other specific embodiments of the present invention, the size of the destination stack 32 and the source stacks 22 to 26 is equal to the bandwidth of the lower destination region 31, namely defined by the cut-off frequency FC.
According to one embodiment of the present invention, one of the source stacks 22 to 26 is selected and transposed to the destination stack 32 by replacing the original frequency content in the destination stack 32 by the frequency content of the selected source stack 22 to 26.
According to a further embodiment of the present invention, one of the source stacks 22 to 26 is selected and transposed to the destination stack 32 by combining the original frequency content in the destination stack 32 and the frequency content of the selected source stack 22 to 26.
According to a still further embodiment of the present invention, a stack-sized frequency area formed out of the source stacks 22 to 26 is selected and transposed to the destination stack 32 by replacing the original frequency content in the destination stack 32 by the frequency content of the newly formed stack-sized frequency area.
According to yet another embodiment of the present invention, a stack-sized frequency area formed out of the source stacks 22 to 26 is selected and transposed to the destination stack 32 by combining the original frequency content in the destination stack 32 with the frequency content of the newly formed stack-sized frequency area.
Generally and particularly regarding the above-mentioned embodiments, frequency components below the cut-off frequency FC remain unchanged.
The frequency stacking algorithm can be generalized, for example, by choosing source and destination stack sizes as a function of the bandwidth defined by the cut-off frequency FC instead of being equal to it. For example, the bandwidth of the source and destination stack sizes may be defined by 0.7, 1.5 or 2 times the cut-off frequency FC.
The combination of frequency content of the source stack or source stacks 22 to 26 with those in the destination stack 32 is done, for example, with a peak picking algorithm.
The frequency stacking algorithm also provides a frequency transposition scheme framework, in which intelligent adaptive frequency transposition can be conveniently implemented and in which most significant spectral segments can be specifically targeted for transposition. For example, such frameworks are described in connection with
In connection with
In the static frequency stacking algorithm, with x being a frequency bin and xε[1, . . . , stack size], each xth frequency bin of the destination stack 32 is replaced by the maximum of the corresponding xth frequency bins of all predefined source stacks 22 to 26. It is noted that in this frequency stacking algorithm the magnitude order of the frequency bins in the destination stack 32 is not necessarily the same as in the original frequency bins in the source region 20. However, this can be managed to a certain extent by applying a weighting function (that is yet to be described) before a transposition step together with a peak picking algorithm to choose between the corresponding frequency bins of the source stacks.
In the example of
The values for the spectral energy SE at the output frequency bins b′, c′ and d′ are calculated similarly. Accordingly, the value for the spectral energy SE at the output frequency b′ is equal to the value for the spectral energy SE at the input frequency bin b of the fourth source stack 25 (arrow B′ in
In
In one embodiment, a source stack 50 is defined around a maximum frequency bin (also called center frequency bin) lying within a stack frequency range comprising all source stacks 22 to 26, for example, the center frequency having maximum spectral energy. In
In a further embodiment, a source stack to be transposed is equal to one of the predefined source stacks 22 to 26. The source stack to be transposed comprises the frequency bin having the maximum spectral energy. As can be seen from
In a still further embodiment, one of the source stacks is selected and transposed to the destination stack. Thereto, the overall energy of the frequency bins pertaining to the same source stack is calculated for each of the predefined source stacks. The predefined source stack with the highest energy sum is then transposed to the destination stack. This is further illustrated in
Yet another possible embodiment of a transposition scheme is one that selects the source stack preserving the maximum spectral contrast.
The present invention offers the opportunity for more intelligent signal processing in a frequency transposition scheme and opens the possibility of a more targeted frequency transposition. This allows for reducing the frequency transposition edge below what is possible with known techniques. It prevents the distortion of vowels which are seen to occur with known transposition schemes on using very low cut-off frequencies FC. The frequency stacking framework in the frequency domain also allows for adaptive frequency transposition by lowering perceptually significant, contiguous chunks of spectral segments or stacks above the cut-off frequency FC. In this respect, the dynamic stacking approaches described herein outperform all known frequency transposition techniques.
The peak picking algorithm, when used in conjunction with a weighting function (yet to be described) and the frequency stacking scheme, allows a convenient second degree of control on what can be transposed, thus allowing for “biased” and “adaptive” frequency transposition for the first time. Such a possibility did not exist, particularly not in known non-linear frequency compression schemes.
The weighting function (also called expectation bias) is used to adaptively choose (or select) different parts of the input spectrum to transpose to the destination region. The spectral energy magnitudes are multiplied by the weights of the weighting function and this weighted spectrum is used to select a particular source stack. Un-weighted signal components of the selected source stack are then processed further i.e. transposed to the corresponding destination stack.
The weighting function or the expectation bias function weights the input spectrum in such a way that the already available low frequency information is given more significance. If a frequency transposition scheme then selects the most important information from a given source region to be transposed to a destination region, auditory expectations are respected more and information is transposed only if it is considerably significant in comparison to what is already accessible to the hearing impaired user in the lower frequencies or destination region.
An advantage of using a weighting function is that an adaptive lowering can be accomplished without any explicit real time detection of phonemes themselves. This is accomplished by a careful choice of weights and by exploiting the fact that fricatives have proportionally much larger energy in the higher frequencies compared to vowels. This keeps the vowels from getting distorted while still lowering high frequency information in fricatives.
The weighting by the weighting function w and the transposition of the source region to the destination region will be explained in detail along with
The speech spectral energy of a human being is distributed across different frequency bands with the difference in distribution corresponding to the different phonemes: vowels, consonants, fricatives, etc. For a given hearing loss with dead regions, i.e. with frequency bands wherein no acoustic perception is possible, high frequency components pertaining in such dead regions (also called source regions hereinafter) are transposed to or shifted to frequency region (also called destination regions hereinafter), in which acoustic perception is still possible. According to the present invention, the destination region is determined, for example, by the hearing loss itself. However, the source region, i.e. important acoustic information that lies in the inaccessible high frequency range, is not fixed but varies with phonemes.
According to the present invention, an adaptive frequency transposition scheme is proposed which reaches a decision for a given spectral energy distribution in the input spectrum. The decision involves choosing the best source frequency range from where energy needs to be transposed to the destination region, and whether to transpose anything at all depending on the energy distribution in the destination region.
Therewith, it is ascertained that the new synthesized sound is as close to the otherwise previously accessible sound to the hearing impaired, or in other words respects the auditory expectations of the hearing impaired user in the best possible way, while still making available the maximum possible new information for enhanced speech comprehension.
Therewith, the present invention helps to minimize initial objections of a hearing device user and helps to reduce acclimatization to the new algorithm. Furthermore, the present invention is a simple solution that can be implemented, for example, by a spectral weighting function to be described below.
The destination region of the frequency transposition scheme can be defined by taking into account a given hearing loss of the user of the hearing device. The source region is assumed to be variable depending on the energy/information distribution, in particular resulting of phonemes. Once the source region and the destination region or destination regions are specified, a comparison/selection scheme being an adaptive algorithm itself is used to choose and process the transposed signal for perceptual benefits. A selective processing may be, for example, a loudness scaling to preserve naturalness of the lowered speech with respect to phonemes/vowels that are only affected in a minor manor by the frequency transposition scheme.
According to one embodiment of the present invention, it is proposed to use a weighting function (also called expectation bias) to adaptively choose or select different parts of the input spectrum to transpose to the same destination region. The spectral energy magnitudes are multiplied by the weights of the weighting function w and this weighted spectrum is used to by a frequency transposition scheme for further processing. In a further embodiment of the present invention, the weighting function w is only applied in order to select a source region. The step of transposing the selected source region is applied to the un-weighted spectrum.
The weighting function w or the expectation bias function weights the input spectrum in such a way that the already available low frequency information is given more significance. If a frequency transposition scheme then selects the most important information from a given source region to be transposed to a destination region, auditory expectations are respected more and information is transposed only if it is considerably significant in comparison to what is already accessible to the hearing impaired user in the lower frequencies or destination region.
An advantage of using a weighting function w is that an adaptive lowering can be accomplished without any explicit real time detection of phonemes themselves. This is accomplished by a careful choice of weights and by exploiting the fact that fricatives have proportionally much larger energy in the higher frequencies compared to vowels. This keeps the vowels from getting distorted while still lowering high frequency information in fricatives.
In principle the present invention can be extended for a low frequency hearing loss as well—although a low frequency hearing loss is rare but still well known—, where the auditory expectation bias or weighting function w is derived from accessible high frequencies.
In
As can be seen from
The diagram of
log(w*x)=log w+log x
Applying the weighting function w to the first spectral contour S1 therefore results in a selection for transposition of the corresponding spectral section (in
The weighting function w of
The purpose of the weighting function w is to alter the significance of the spectral information based on expectation of the hearing impaired person.
This significance measure which is obtained by multiplication of the weighting function w by the spectral energy magnitude—represented by the spectral contours S1 and S2 in
Since most hearing impaired patients are still having some usable hearing in the low frequency section L, a weighting function w according to the present invention can be chosen such that a lot of importance is given to low frequency information to keep them from getting modified (and distorted) by a frequency transposition scheme. The method according to the present invention proposes to bias a frequency transposition scheme to better match the auditory expectation of a hearing impaired user based on available hearing, whereas still leaving the door open for transposing fricatives dominated by high frequency energies. Together with a weighting function w, a frequency transposition scheme can exploit the fact that vowels are dominated by higher energies in lower frequencies, and fricatives by higher energies in the higher frequencies, to conditionally lower fricatives while leaving vowels almost untouched. This decreases the initial objections of the hearing impaired user to a frequency transposition scheme while maximizing benefit. This could be critical to acceptance of a frequency transposition algorithm for those hearing impaired users, where a cut-off frequency needs to be reduced low enough to encroach on the frequency area with significant vowel information, i.e. lower than 1′500 Hz. An assumption that the frequency transposition scheme makes is that the energy distribution of the phonemes is not significantly different for the contiguous frequencies in the most important spectral section of the phoneme, which appears to be reasonable. This is a prerequisite so that the most significant spectral section thrown up after applying the weighting function w is perceptually coherent and meaningful.
One can see in
The weighting function w, w′ described here can be used with all frequency transposition schemes, be it for speech or for music. However, the success of the frequency transposition will depend on the frequency transposition scheme itself. In particular, a piecewise division of the input spectrum—at least into a source region and a destination region—is important for a meaningful selection of the frequency section preferred for the transposition. Even in frequency transposition schemes that use linear frequency transposition, the proposed weighting functions w, w′ can be used to protect important spectral information in the destination region from getting disrupted. One of the advantages of achieving adaptive lowering with this kind of weighting function w, w′ is the ease of integration with the frequency transposition scheme itself. For example, the simple weighting functions w shown in
The actual values for the weight function w, w′ can be used to more specifically target a given phoneme for frequency transposition in order to arrive at a trade-off between sound quality and benefit of transposed information. It is further to be noted that the simple weighting scheme described here is approximate and not exact in the sense that it just offers an easily parameterizable trade-off in a frequency transposition context between what can be transposed and the distortions that can still be tolerated, to arrive at an optimal fitting of a frequency transposition scheme for a given hearing loss.
The frequency transposition scheme described as a possible embodiment for a frequency lowering is called frequency stacking and has been extensively described in connection with the embodiments depicted in
The frequency transposition scheme described in connection with
The frequency transposition scheme illustrated in
The two weighted spectral contours WS corresponding to a vowel (
The presented weighting functions w, w′ can be used in a frequency transposition scheme to push the cut-off frequency FC further down than it is possible with state of the art algorithm. It could potentially be used in all hearing devices that use a frequency compression and where it makes sense to offer lower cut-off frequencies for a sound recover feature while managing the adaptation time and/or initial objections by hearing device users.
In
Again, the input frequency fin is shown on the x-axis while the output frequency fout is shown on the y-axis. The x-axis as well as the y-axis has a logarithmic scale. The frequency transposition scheme according to the present invention comprises the step of copying the spectral energy in the lower source region 21 to the lower destination region 31 up to the lower cut-off frequency FC (one-to-one mapping). Furthermore, the spectral energy of a first source stack 22, which starts at the lower cut-off frequency FC and ends at a upper cut-off frequency FHL, is—in one embodiment—also copied to a destination stack 32 (again one-to-one mapping). While the upper cut-off frequency FHL is ideally envisaged to be the edge of the aid-able region of hearing for the hearing device user (it is noted that the upper cut-off frequency FHL could also be higher or lower than the edge of the aid-able region of hearing), up to which upper cut-off frequency FHL the auditory expectations of the hearing device user need to be respected, the lower cut-off frequency FC is determined by the following equation:
wherein
The determination of the optimal values of parameters in the above equation for a given hearing loss could be based on audiological experiments that are described, for example, in a publication entitled “Modified Verification Approaches for Frequency Lowering Devices” by Danielle Glista & Susan Scollie (National Centre for Audiology, the University of Western Ontario, Sep. 11, 2009). This publication can be retrieved from the internet under http://www.audiologyonline.com/articles/article_detail.asp? article_id=2301.
In the frequency transposition scheme according to the present invention, the compression does not start at the lower cut-off frequency FC but at the upper cut-off frequency FHL. The compression ends at the upper frequency Fu, above which no relevant information is expected. The second source stack 23—defined between the upper cut-off frequency FHL and the upper frequency Fu—is transposed as well to the destination stack 32, in which a replacement and/or superposition of spectral energy of the first source stack 22 and/or the second source stack 23 takes place. For example, a biased peak picking algorithm or a weighting function w with subsequent superposition is applied to emphasize relevant spectral information in the second source stack 23 or in the first source stack 22.
The biased peak picking method is used to respect the auditory expectation of the hearing device user and is achieved by using an appropriate spectral weighting function.
The weighting function w (again also called expectation bias) is used to adaptively choose different parts of the input spectrum—e.g. the first source stack 22 or the second source stack 23 (
The weighting function w or the expectation bias function weights the input spectrum in such a way that the already available low frequency information is given more significance. If a frequency transposition scheme then selects the most important information from a given source region 20 to be transposed to a destination region 30, auditory expectations are respected more and information is transposed only if it is considerably significant in comparison to what is already accessible to the hearing impaired user in the lower source region 21 or the lower destination region 31.
An advantage of using a weighting function w is that an adaptive lowering can be accomplished without any explicit real time detection of phonemes themselves. This is accomplished by a careful choice of weights and by exploiting the fact that fricatives have proportionally much larger energy in the higher frequencies compared to vowels. This keeps the vowels from getting distorted while still lowering high frequency information in fricatives.
The frequency transposition scheme according to this embodiment of the present invention ensures two things which are fundamentally different from the known frequency compression scheme:
First, it separates the second source stack 23 from the first source stack 22 in the frequency transposition context. The second difference is that the final output of the frequency transposition scheme in the destination stack 32 is chosen with a biased peak picking algorithm between the spectral energies of the first source stack 22 and the second source stack 23. This results in the final input/output curve becoming signal dependent unlike in the known frequency transposition scheme where a non-linear monotonic relationship between the input frequency fin and the output frequency fout, as shown in
For example, the biased peak picking algorithm can be formulated as follows, wherein for a given compression ratio CR and lower cut-off frequency FC the frequency bins at (n+j), n+j+1), . . . , (n+j+k) map to the same frequency bin n:
i=arg max [w(n)·Fin(n),w(n+j)·Fin(n+j),w(n+j+1)·Fin(n+j+1), . . . ,w(n+j+k)·Fin(n+j+k)]
Fout(n)=w″(i)·Fin(i)
wherein
The lower cut-off frequency of the frequency transposition scheme according to the present invention is denoted by FC and the compression ratio applied in the second source stack 23 is denoted by CR.
The relationship between different parameters shown in
The parameterization of the lower cut-off frequency FC and the compression ratio CR in the known frequency compression algorithm should ideally be dependent on the hearing loss and spectral energy distribution of speech.
The separation of the second source stack 23 and the destination stack 32 in the compression scheme, together with a biased peak picking allows for transposing energies only when they are significant compared to what is already there in the first source stack 22. This leaves the already audible harmonic structure of the vowels intact while still transposing fricatives and other phonemes dominated by high frequency energies.
As the harmonic relationship of the notes of western instrumental music is similar to vowels, the frequency transposition scheme according to the present invention also distorts music less in comparison to the known techniques.
All embodiments of the present invention allow to apply frequency transposition schemes to be extended to hearing impaired with profound hearing losses and a very limited bandwidth of aid-able hearing, by better managing the vowel distortions audible with lower cut-off frequencies in the original frequency compression scheme.
In principle the present invention can be extended for a low frequency hearing loss as well—although a low frequency hearing loss is rare but still well known—, where the auditory expectation bias or weighting function is derived from accessible high frequencies.
| Filing Document | Filing Date | Country | Kind | 371c Date |
|---|---|---|---|---|
| PCT/EP2011/060541 | 6/23/2011 | WO | 00 | 12/20/2013 |
| Publishing Document | Publishing Date | Country | Kind |
|---|---|---|---|
| WO2012/175134 | 12/27/2012 | WO | A |
| Number | Name | Date | Kind |
|---|---|---|---|
| 7248711 | Allegro et al. | Jul 2007 | B2 |
| 20040175012 | Roeck et al. | Sep 2004 | A1 |
| Number | Date | Country |
|---|---|---|
| 1441562 | Jul 2004 | EP |
| 1 686 566 | Aug 2006 | EP |
| 2 375 782 | Oct 2011 | EP |
| 2007000161 | Jan 2007 | WO |
| Entry |
|---|
| Written Opinion for PCT/EP2011/060541 dated dated Apr. 10, 2012. |
| International Search Report for PCT/EP2011/060541 dated Apr. 10, 2012. |
| Number | Date | Country | |
|---|---|---|---|
| 20140105435 A1 | Apr 2014 | US |
