EP3232684B1 - Method for transmitting an audio signal from a transmitter to a receiver - Google Patents

Description

The invention relates to a method for transmitting an audio signal from a transmitter to a receiver. The invention further relates to a hearing aid and a hearing aid system with two such hearing aids. The hearing aid is preferably a hearing aid.

Persons suffering from a reduction in hearing usually use a hearing aid. In this case, an ambient sound is usually detected by means of an electromechanical sound transducer. The detected electrical signals are processed by means of an amplifier circuit and introduced by means of another electromechanical transducer in the ear canal of the person. Different types of hearing aids are known. The so-called "behind-the-ear devices" are worn between the skull and the auricle. The introduction of the amplified sound signal into the ear canal takes place here by means of a sound tube. Another common embodiment of a hearing aid is an "in-ear device" in which the hearing aid itself is inserted into the ear canal. Consequently, the auditory canal is at least partially closed by means of this hearing aid, so that no further sound - or only a greatly reduced amount of sound - can penetrate into the auditory canal, apart from the sound signals generated by the hearing aid.

If the person suffers from a hearing impairment in both ears, a hearing aid system with two such hearing aid is used. Here, each of the ears is assigned to one of the hearing aids. In order to allow the person a spatial hearing, it is necessary that with a the hearing aids detected audio signals are provided to each other hearing aid device. On the one hand, this requires transmission with only a comparatively small time offset. On the other hand, the head of the person acts as an attenuation, which is why the transmission rate between the hearing aids is limited. In addition, because of the limited energy storage of hearing aids and the otherwise excessive burden on the person a transmission power limited.

The invention has for its object to provide a particularly suitable method for transmitting an audio signal from a transmitter to a receiver and a particularly suitable hearing aid and a particularly suitable hearing aid system with two hearing aids, in particular an audio quality is improved, and preferably reduces a transmission rate is. According to the invention, this object is achieved with regard to the method by the features of claim 1 and with regard to the hearing device by the features of claim 7 and with respect to the hearing aid system by the features of claim 8. Advantageous developments and refinements are the subject of the respective subclaims.

The method is for transmitting an audio signal from a transmitter to a receiver, wherein the transmitter or the receiver is preferably a component of a hearing device. The respectively remaining element, that is to say the transmitter or the receiver, is suitably a component of a further component of a hearing device system having the hearing device.

For example, the hearing aid is a headphone or includes a headphone. However, the hearing aid is particularly preferably a hearing aid. The hearing aid is used to support a person suffering from a reduction in hearing. In other words, the hearing aid is a medical device by means of which, for example, a partial hearing loss is compensated. The hearing aid is, for example, a "receiver-in-the-canal" hearing aid (RIC), an in-the-ear hearing aid, such as an "in-the-ear" hearing aid. an in-the-canal hearing aid (ITC) or a complete-in-canal hearing aid (CIC), a pair of hearing glasses, a pocket hearing aid, a bone conduction hearing aid or an implant. The hearing aid device is particularly preferably a behind-the-ear hearing aid device ("behind-the-ear" hearing aid device) which is worn behind an auricle.

The method provides that, at the transmitter end, an input signal corresponding to the audio signal is temporally divided into time windows, wherein the length of the time windows is preferably the same. The length of the time window is, for example, between 0.5 ms and 2 ms and in particular equal to 1 ms. The input signal is preferably the audio signal or in part thereof. For example, the audio signal is decomposed into different input signals, each input signal being respectively divided into different time windows, which differ in particular on the basis of their length. For a given time window, the input signal is divided on the transmitter side into a number of frequency channels. On the transmitter side, each frequency channel is assigned a current channel value. The channel value is an amplitude and / or a phase or a signal level. On the transmitter side, the current channel values are divided into a first current data record and a second current data record, wherein the first current data record and the second current data record each comprise at least one of the current channel values. In particular, the first current data record has only a single one of the current channel values.

On the basis of at least one chronologically preceding channel value, a first prognosis for the first current data record is created on the transmitter side. The creation takes place in particular such that a difference ("pädikationsfehler") between the first current record and the first prognosis is as small as possible. Conveniently, the first forecast includes as many values as the first current record. For example, the first forecast includes only a single value. For example, the temporally preceding channel value has been assigned to one of the frequency channels one time slot earlier, for example the same frequency channel for which the prognosis is being compiled, in particular if the first current record comprises only a single one of the current channel values. Preferably, a number of temporally preceding channel values are used, whereby, for example, channel values from other frequency channels and / or channel values are used whose time interval varies.

On the transmitter side, a first deviation between the first prognosis and the first current dataset is determined, ie by how much the first prognosis and the first current dataset differ. Furthermore, on the transmitter side, based on the first current data record, a second prognosis for the second current data record is created. In this case, for example, the first current data record and / or the first deviation are used directly. The second prognosis expediently comprises just as many values as the second current dataset, wherein the generation of the second prognosis takes place in particular such that a difference between the second current dataset and the second prognosis is as small as possible. On the transmitter side, a second deviation between the second prognosis and the second current dataset is determined, ie by how much the second prognosis and the second current dataset differ.

A first transmission value corresponding to the first deviation is transmitted from the transmitter to the receiver, wherein the first transmission value is expediently first created on the transmitter side based on the first deviation. The first transmission value preferably has a lower dimensionality or at most the same dimensionality as the first deviation, and is for example a one-dimensional value. A second transmission value corresponding to the second deviation is transmitted by the transmitter to the receiver, wherein the second transmission value is expediently first created on the transmitter side on the basis of the second deviation. The second transmission value preferably has a lower dimensionality or at most the same dimensionality as the second deviation, and is for example a multidimensional or particularly preferably a one-dimensional value. Preferably, the dimensionality of the first and second transmission values is the same.

In a further step, a first reconstructed data record is created on the receiver side on the basis of a previously reconstructed output signal and the transmitted first transmission value. As a temporally preceding reconstructed output signal, it is expedient to use a value or a data record which corresponds to the temporally preceding channel value present at the transmitter end. For example, zero (0) is used at the beginning of the method as the temporally preceding channel value or as a previously reconstructed output signal, or a specific value is first transmitted from the transmitter to the receiver, and both at the transmitter and at the receiver as temporally preceding channel value or used as temporally preceding reconstructed output signal. The first reconstructed data record created in this way thus essentially corresponds to the first data record on the transmitter side, with differences preferably being present only on the basis of the generation of the first transmission value.

In a further step, a second reconstructed data record is created on the receiver side based on the first reconstructed data record and the transmitted second transmission value. In particular, additional data that are available on the receiver side are used for the creation. The second reconstructed data record created in this way essentially corresponds to the second data record present on the transmitter side, with differences preferably being present only on the basis of the generation of the second transmission value and / or due to differences between the first reconstructed data record and the first data record.

The first reconstructed data set and the second reconstructed data record are combined on the receiver side into a reconstructed output signal. The summary here is in particular the inverse function for the transmitter-side distribution of the current channel values into the first current data record and the second current data record, so that the transmitter-side distribution of the input signal into the frequency channels is present on the receiver side. The reconstructed output signal essentially corresponds to the transmitter side existing distribution of the input signal in the frequency channels. For example, on the receiver side, the reconstructed output signal is further processed and the individual frequencies are combined and transferred into the time domain. For example, when the method is executed again, the reconstructed output signal is used and used as the previously preceding reconstructed output signal. Preferably, the method is executed again after the expiry of the specific time window.

Preferably, the first transmission value is created as soon as the first deviation is created, and then transmitted directly thereafter to the receiver, in particular temporally before or at least simultaneously before the creation of the second prognosis. In particular, the second prediction / the second deviation and the first reconstructed data record are essentially created at the same time, so that a simultaneous processing takes place on the transmitter and receiver side, which is why a transmission speed is increased.

The second prognosis is expediently also generated on the transmitter side on the basis of the temporally preceding channel value and the second reconstructed dataset on the receiver side based on the temporally preceding reconstructed output signal, wherein, for example, a number of temporally preceding reconstructed output signals or temporally preceding channel values are used. In this way, the second deviation is reduced.

Preferably, a linear prediction is used to generate the first prognosis and the first reconstructed dataset. Alternatively or in combination with this, a linear prediction is used to generate the second prognosis and the second reconstructed dataset. In other words, each value is created by means of a linear combination, wherein preferably a number of temporally preceding channel values, a number of temporally preceding reconstructed output signals, a number of first current data records or a number of first reconstructed data records are used.

oder $$\widehat{\mathbf{x}}\left(\mathrm{n}\right)={\displaystyle \sum _{\mathrm{i}=1}^{\mathrm{N}}\mathbf{Ay}\left(\mathrm{n}-\mathrm{i}\right)}$$

ermittelt. x̂(n) bezeichnet hierbei die ersten Prognose, den ersten rekonstruierten Datensatz, die zweiten Prognose bzw. den zweiten rekonstruierten Datensatz. a_i bezeichnet einen Koeffizienten, A eine Koeffizientenmatrix und y die Gesamtheit der Werte, die zur Erstellung herangezogen werden, insbesondere die zeitlich vorhergehenden Kanalwerte, die zeitlich vorhergehenden rekonstruierten Ausgangssignale, die ersten aktuellen Datensätze bzw. die ersten rekonstruierten Datensätze, wobei insbesondere eine Anzahl derartiger Werte herangezogen wird, deren jeweiliger Erstellungszeitpunkt sich unterscheidet. Der Erstellungszeitpunkt ist hierbei n - i, und die verwendete Anzahl ist N. Eine Art von linearer Vorhersage ist beispielsweise in " Benesty, J., Chen, J., & Huang, Y. (Arden). (2008). Linear Prediction. In J. Benesty, M. M. Sondhi, & Y. (Arden) Huang (Eds.), Springer Handbook of Speech Processing (pp. 111-125). Springer Verlag " offenbart, insbesondere in Kapitel 7.2 (Seite 112 -113), insbesondere Formel 7.6, sowie insbesondere in Kapitel 7.9 (Seite 120 -124), insbesondere Formel 7.108.In summary, the values of the first prognosis, the first reconstructed dataset, the second prognosis or the second reconstructed dataset are calculated using the formula $$\widehat{\mathbf{x}}\left(\mathrm{n}\right)={\displaystyle \underset{\mathrm{i}=1}{\overset{\mathrm{N}}{\Sigma }}{\mathrm{a}}_{\mathrm{i}}\mathrm{y}\left(\mathrm{n}-\mathrm{i}\right)}$$

or $$\widehat{\mathbf{x}}\left(\mathrm{n}\right)={\displaystyle \underset{\mathrm{i}=1}{\overset{\mathrm{N}}{\Sigma }}\mathbf{Ay}\left(\mathrm{n}-\mathrm{i}\right)}$$

determined. x (n) denotes the first prognosis, the first reconstructed dataset, the second prognosis or the second reconstructed dataset. a _i denotes a coefficient, A a coefficient matrix and y the totality of the values used for the generation, in particular the temporally preceding channel values, the temporally preceding reconstructed output signals, the first current data sets or the first reconstructed data records, in particular a number of such Values are used, the respective date of their creation differs. The creation time here is n - i, and the number used is N. One kind of linear prediction is, for example, in " Benesty, J., Chen, J., & Huang, Y. (Arden). (2008). Linear Prediction. In J. Benesty, MM Sondhi, & Y. (Arden) Huang (Eds.), Springer Handbook of Speech Processing (pp. 111-125). Springer Verlag ", in particular in chapter 7.2 (pages 112-113), in particular formula 7.6, and in particular in chapter 7.9 (pages 120-124), in particular formula 7.108.

For example, the input signal is divided into the frequency channels by means of a Fourier transformation. However, particularly preferred channel filter are used, which are preferably combined in a filter bank. For example, the first transmission value is created by quantizing the first deviation, and / or the second transmission value is created by quantizing the second deviation. Here, the first deviation of the first transmission value is assigned, which expediently only can assume a discrete number of different values. The second deviation is assigned the second transmission value, which can expediently assume only a discrete number of different values, the number being for example different or equal to the number of possible values of the first transmission value. In other words, the first or second transmission value is a discrete value.

On the transmitter side, a third reconstructed data set is preferably generated on the basis of the first transmission value and / or the second transmission value and on the basis of the first prognosis or second prognosis. The third reconstructed data set corresponds to the reconstructed output signal but is present on the transmitter side. In other words, on the transmitter side, the output signal is also reconstructed on the basis of the first transmission value or the second transmission value, whereby the output signal may deviate slightly from the input signal due to the quantization and the induced noise introduced as a result. The third reconstructed data set is used as temporally preceding channel value. At least one of the values of the third reconstructed data set or all values for this is used. In this way, deviations between the output signal and the input signal due to the quantization are taken into account in the preparation of the first / second prognosis, which is why a maximum deviation between the input signal and the reconstructed output signal remains low even with a repeated execution of the method Quality when transmitting the audio signal is present. In particular, the advantage of this method is that only information / signals are used that are available both at the transmitter and at the receiver. Thus, the reconstructed signals in the transmitter and receiver are identical (at least in the case of faultless transmission).

Suitably, for the quantization of the first deviation, the same quantization is used as for the quantization of the second deviation. For example, a scalar quantization is used. More preferably, the quantization is a vector quantization. Suitably, a so-called gain-shape vector quantization is used. The quantized signal is here divided into the signal shape / vector shape and a scaling factor (gain). A particularly suitable form of gain-shape vector quantization is the logarithmic vector quantization, in particular the (spherical) logarithmic vector quantization. In this case, possible signal forms / vector shapes are points on a (potentially) high-dimensional unit sphere (ie with radius 1). The scaling factor is quantized logarithmically, for example, with the known A-law. Waveforms / vector shapes may also be other shapes such as (high dimensional) pyramids or cubes. For example, spherical-logarithmic vector quantization is " B. Matschkal and JB Huber, "Spherical Logarithmic Quantization", IEEE Trans. Audio, Speech, and Language Processing, vol. 18, pp. 126-140, Jan. 2010 in particular from Chapter III, an example being disclosed in Chapter IV, in particular in Figs. 8 and 9.

Preferably, the first reconstructed data record is created by using the first transmission value to create a first auxiliary data record which corresponds to the first deviation. In other words, in an application of the same creation rule, on the basis of which the first transfer value is created, to the first auxiliary data record, the first transfer value would result. In particular, a one-to-one function is used to generate the first transmission value, and the inverse function of this is used to generate the first auxiliary data set on the basis of the first transmission value. Furthermore, based on the temporally preceding reconstructed output signal, a first auxiliary prognosis is created, which corresponds to the first prognosis, and this preferably corresponds. Preferably, the same calculation rule is preferably used to generate the first auxiliary prognosis on the basis of the previously reconstructed output signal, as for the generation of the first prognosis on the basis of the temporally preceding channel value. Conveniently, the temporally preceding reconstructed output signal is also divided into channels corresponding to the frequency channels. The first auxiliary data set is added to the first auxiliary prognosis. Suitably, the addition takes place in a value-wise manner. In other words, corresponding values of the two data sets are added together in each case, and this sum forms a value in each case of the first reconstructed data set. In summary, each value of the first auxiliary data set is added to a value of the first auxiliary prognosis. In particular, the first reconstructed record corresponds to the first current record. In other words, all values of the two data sets are substantially the same.

Preferably, the second reconstructed data record is created by using the second transmission value to create a second auxiliary data record which corresponds to the second deviation. In other words, in an application of the same creation rule, on the basis of which the second transfer value is created, to the second auxiliary data record, the second transfer value would result. In particular, a one-to-one function is used to generate the second transmission value, and the inverse function thereof is used to generate the second auxiliary data set on the basis of the second transmission value. Furthermore, based on the first reconstructed data set, a second auxiliary prognosis is created, which corresponds to the second prognosis, and this preferably corresponds. Preferably, to generate the second auxiliary prognosis, the same calculation rule is preferably used on the basis of the first reconstructed data set as for the generation of the second prognosis on the basis of the first data set. The second auxiliary data set is added to the second auxiliary prognosis. Suitably, the addition takes place in a value-wise manner. In other words, the corresponding value of the two data sets is added together in each case, and this sum forms in each case a value of the second reconstructed data record. In summary, each value of the second auxiliary data set is added to a value of the second auxiliary prognosis. In particular, the second reconstructed record corresponds to the second current record. In other words, all values of the two data sets are substantially the same.

If the third reconstructed data record is used, preferably both the first and the second auxiliary data record are created on the transmitter side and in each case added to the values of the first or second prognosis. If, as a result of the quantization, noise is introduced into the first or second transmission value, this is determined when the first one is determined again or second prognosis or auxiliary prognosis, both on the transmitter and on the receiver side.

Preferably, the first and / or second deviation is created by creating the difference between each current channel value of the first current data set or the second current data set and an associated prediction value of the first or second prediction to form a difference value. In other words, each current channel value of the respective current data record is subtracted from a corresponding forecast value of the respective prognosis, or each forecast value of the respective prognosis is subtracted from the corresponding current channel value of the respective current data record. The difference values created here form the first or second deviation. In other words, the first / second deviation is a data set / vector, and the number of difference values of the first deviation is equal to the number of current channel values of the first current data set equal to the number of prediction values of the first prediction. Also, the number of difference values of the second deviation is equal to the number of the current channel value of the second current data set and equal to the number of forecast values of the second prognosis. The generation of the individual forecast values of the first prognosis takes place, in particular, independently of one another, and the prognosis values are preferably created at least partially in parallel over time. Alternatively or particularly preferably in combination with this, the generation of the individual forecast values of the second prognosis takes place in particular independently of one another, and the prognosis values are preferably created at least partially in parallel over time.

For example, the current channel values are split substantially equally between the first current record and the second current record, so that the first current record thus has substantially the same number of channel values as the second current record, for example, the numbers being one (1) differ. In this case, in particular each frequency channel is assigned an index, wherein expediently all frequency channels with a straight index one of the current data sets and the frequency channels with an odd index the remaining current data set be assigned. For example, all frequency channels with an odd index are assigned to the first current record. As a result of the procedure, in particular the first reconstructed data record and the second reconstructed data record also have essentially the same number of reconstructed channel values.

In summary, in particular, the input signal is decomposed into a number N of frequency channels. This creates a signal that depends on both the time and the frequency channel. The signal thus represents the current channel values and the temporally preceding channel values. The signal can be denoted by x (t, k), where t denotes the (discrete) time index and k the channel index. The totality of the current channel values is thus x (t1, k), where the time index t1 denotes the current time / window. Every x (t, k) is generally a complex number. For a compact mathematical representation of the method, for example, x (t, k <1) = 0. Also, let x (t, k> N + 1) = 0.

M ⁺ ≥ 0, M ^- ≤ 0, I_m ≥ 1, L ≥ 1
"x Dach" bezeichnet hierbei die erste und zweite Prognose ("Prädiktionswert") und "x Tilde" den rekonstruierten Wert/die aktuellen bzw. zeitlich vorhergehenden Kanalwerte. Wenn der zweite Term vorhanden ist, ist "x Dach" beispielsweise der zweiten Prognose, und sonst der ersten Prognose zugeordnet.The first and second forecasts are created in particular using the formula, wherein the current time / the current time window is denoted by t, $$\widehat{x}\left(t,k\right)={\displaystyle \underset{m=M}{\overset{M^{+}}{\Sigma }}{\displaystyle \underset{i=1}{\overset{I_m}{\Sigma }}{a}_{i.m}^k\cdot }}\tilde{x}\left(t-i.k-m\right)+{\displaystyle \underset{\mathcal{l}=1}{\overset{L}{\Sigma }}{a}_{\mathcal{l}}^k}\cdot \tilde{x}\left(t.k-\mathcal{l}\right)$$

M ⁺ ≥ 0, M ^- ≦ 0, I _m ≥ 1, L ≥ 1
"x roof" here denotes the first and second forecasts ("prediction value") and "x tilde" the reconstructed value / the current or temporally preceding channel values. For example, if the second term exists, then "x Roof" is assigned to the second forecast, and otherwise to the first forecast.

The double sum in the above formula makes a contribution to the prediction value based solely on values from previous time steps (due to "ti") (temporally preceding channel values), but not necessarily from the same Channel (but from channel km, where m can also be 0). The fact that the second sum goes up to l_m takes into account the fact that different channels can look "differently" into the past.

Due to the second sum, values from the same time step (current channel values) are also used. In principle, all values from channels that have already been calculated can be used for this purpose. In the above formula, this is done "from bottom to top". The lowest channel (channel number 1) is predicted only from temporally preceding values. This also follows from the equation above due to the zeroing of the (hypothetical) channels with channel numbers <1. In the next step (for channel number 2) channel number 1, ie x tilde (t, 1) can be used, etc. This is conceivable Procedure, of course, also in reverse channel order and in principle even with any channel order. Thus, if the second sum is present, "x roof" preferably designates a forecast value of the second prognosis.

For example, the prediction based on temporally preceding values (the double sum) is performed in parallel in a plurality of frequency channels, and thus the first prediction is made. Here, the frequency channels can be chosen, for example, equidistant, with any selection is possible, such as every other frequency channel. This method has the advantage that the values which can now be calculated in parallel can be quantized by means of vector quantization, which i.a. leads to a lower quantization noise. In a second and following step, the reconstructed values of the same time step obtained in the previous step can now be used for the prediction. In other words, the second forecast is created. For example, the creation is done for all odd channel numbers.

The hearing device has a communication device for transmitting and / or receiving an audio signal. For this purpose, the communication device comprises a transmitter or a receiver. The communication device is suitable as well as provided and set up a method for transmitting an audio signal from the transmitter or to the receiver. In this case, the method provides that, on the transmitter side, an input signal corresponding to the audio signal for a specific time window is divided into a number of frequency channels, and that a current channel value is assigned to each frequency channel on the transmitter side. Furthermore, the current channel values are divided on the transmitter side into a first current data record and a second current data record, and on the transmitter side, a first prognosis for the first current data record is created on the basis of a preceding channel value. In a further step, a first deviation between the first prognosis and the first current dataset is determined on the transmitter side, and a second prognosis for the second current dataset is created on the transmitter side on the basis of the first current dataset. On the transmitter side, a second deviation between the second prognosis and the second current dataset is determined.

Furthermore, the method provides that a first transmission value corresponding to the first deviation and a second transmission value corresponding to the second deviation are transmitted from the transmitter to the receiver. On the receiver side, a first reconstructed data record is created based on a temporally preceding reconstructed output signal and the transmitted first transmission value, and on the receiver side, a second reconstructed data record is created on the basis of the first reconstructed data record and the transmitted second transmission value. In a further working step, the receiver side reconciles the first reconstructed data set and the second reconstructed data record into a reconstructed output signal.

If the communication device only comprises the transmitter, in this case in particular only the transmitter-side working steps and a work step for transmitting the deviations are carried out. If the communication device only has the receiver, in particular only the receiver-side working steps and a work step for receiving the deviations are carried out. Conveniently, the transmission is wireless, such as inductive or wireless.

The hearing device preferably comprises a sensor, by means of which an audio signal is detected during operation. The sensor is preferably an electromechanical transducer, such as a microphone. For example, the input signal is the audio signal or the input signal is generated based on the audio signal. For example, the input signal is a part of the audio signal, or corresponds to a certain frequency range of the audio signal. To generate the input signal from the audio signal, the hearing device comprises, for example, a signal processing unit and / or filter. The hearing device preferably comprises an amplifier circuit by means of which the audio signal can be amplified. Preferably, the hearing aid comprises an actuator, by means of which a sound signal is generated, such as a loudspeaker, and which is suitable for outputting the output signal or the reconstructed output signal, and is provided and suitable, for example.

For example, the hearing aid is a headphone or includes a headphone. However, the hearing aid is particularly preferably a hearing aid. The hearing aid is used to support a person suffering from a reduction in hearing. In other words, the hearing aid is a medical device by means of which, for example, a partial hearing loss is compensated. The hearing aid is, for example, a "receiver-in-the-canal" hearing aid (RIC), an in-the-ear hearing aid such as an in-the-ear hearing aid, an in-the-ear -canal "- a hearing aid (ITC) or a complete-in-canal hearing aid (CIC), a pair of hearing glasses, a pocket hearing aid, a bone conduction hearing aid or an implant. The hearing aid device is particularly preferably a behind-the-ear hearing aid device ("behind-the-ear" hearing aid device) which is worn behind an auricle.

The hearing aid is provided and adapted to be worn on the human body. In other words, the hearing aid preferably comprises a holding device, by means of which an attachment to the human body is possible. If the hearing device is a hearing aid device, the hearing device is provided and set up, for example, to be arranged behind the ear or within an auditory canal. In particular, the hearing aid is wireless and provided and set up, at least partially inserted into an ear canal to become. For example, the hearing aid is a component of a hearing aid system which comprises a further hearing aid or a further device, such as a directional microphone or another device having a microphone. In this case, the device preferably comprises the transmitter and the hearing device the receiver, and the transmission of the audio signal between the transmitter and the receiver is carried out according to the method.

The hearing aid system preferably comprises two hearing aids, each having a communication device, which are provided and arranged for transmitting and / or receiving an audio signal according to the above method. In this case, the hearing device system is suitable as well as provided and set up to transmit audio signals between the two hearing aids by means of their respective communication devices, wherein the above method is performed. In particular, each of the communication devices each has a transmitter and a receiver, and the audio signals are transmitted between the two communication devices, at least from one of the hearing aids to the remaining one. Conveniently, the transmission is wireless, such as inductive or wireless.

The hearing aid system is particularly preferably a hearing aid system. The hearing aid system is used to support a person suffering from a reduction in hearing. In other words, the hearing aid system is a medical device by means of which, for example, a partial hearing loss is compensated. The hearing aid system expediently comprises a behind-the-ear hearing aid that is worn behind an auricle, a "receiver-in-the-canal" hearing aid (RIC), an in-the-ear hearing aid, such as an in-the-ear hearing aid, an in-the-canal hearing aid (ITC) or a complete-in-canal hearing aid (CIC), a pair of hearing-glasses, a pocket-hearing aid, a bone conduction hearing aid or but an implant. The hearing aid system is in particular provided and adapted to be worn on the human body. In other words, the hearing aid system preferably comprises a holding device, by means of which an attachment to the human body is made possible. If the hearing aid system is a hearing aid system, then at least one of the hearing aids provided and set up, for example, to be arranged behind the ear or within an ear canal. In particular, the hearing aid system is wireless and designed and arranged to be at least partially inserted into an ear canal. Particularly preferably, the hearing device system comprises an energy store, by means of which a power supply is provided.

The further developments and advantages described in connection with the method are also to be transferred analogously to the hearing device or the hearing device system and vice versa.

Fig. 1

schematisch ein Hörgerätesystem mit zwei Hörgeräten,

Fig. 2

ein Verfahren zum Übertragen eines Audiosignals zwischen den beiden Hörgeräten,

Fig. 3

ein zu dem Audiosignal korrespondierendes Eingangssignal, und

Fig. 4 - 6

jeweils teilweise Datensätze.

Embodiments of the invention will be explained in more detail with reference to a drawing. Show:

Fig. 1

schematically a hearing aid system with two hearing aids,

Fig. 2

a method for transmitting an audio signal between the two hearing aids,

Fig. 3

an input signal corresponding to the audio signal, and

Fig. 4-6

each partial records.

Corresponding parts are provided in all figures with the same reference numerals.

In Fig. 1 a hearing aid system 2 is shown with two identical hearing aids 4, which are provided and adapted to be worn behind an ear of a user. In other words, it is in each case behind-the-ear hearing aids ("behind-the-ear" - hearing aid), which has a sound tube, not shown, which is inserted into the ear. Each hearing aid device 4 comprises a housing 6, which is made of a plastic. Within the housing 6, a microphone 8 with two electromechanical transducers 10 is arranged. By means of the two electromechanical transducers 10, it is possible to change a directional characteristic of the microphone 8 by a temporal offset between the means of the respective electromechanical transducer 10 detected acoustic signals is changed. The two electromechanical sound transducers 10 are signal-coupled to a signal processing unit 12, which comprises an amplifier circuit. The signal processing unit 12 is formed by means of circuit elements, such as electrical and / or electronic components.

Furthermore, a loudspeaker 14 is signal-wise coupled to the signal processing unit 12, by means of which the audio signals 16 received by the microphones 8 and / or processed by the signal processing unit 12 are output as sound signals. These sound signals are conducted by means of a sound tube not shown near the ear of a user of the hearing aid system 2.

Each of the hearing aid devices 4 also has a transmitter 18 and a receiver 20, by means of which an exchange of data signals 22 between the two hearing aids 4 takes place. The exchange takes place, for example, by radio or inductively. The signal processing unit 12, the transmitter 18 and the receiver 20 in each case essentially together form a communication device 24. Due to the exchange of the data signals 22, it is possible for the wearer of the hearing device system 2 to convey a spatial sense of hearing. In summary, the hearing aid system 2 is configured binaurally.

In Fig. 2 a method 26 is shown, according to which the audio signals 16 are transmitted between the two hearing aids 4 by means of their respective communication device 24. In a first step 28, the audio signal 16 is received by means of a hearing aid 4. In a subsequent second step 30, an input signal 32 is created from this, which consequently corresponds to the audio signal 16 and corresponds to the input signal 32 in FIG FIG. 3 is shown by way of example. For this purpose, the audio signal 16 is filtered, for example. Furthermore, the input signal 32 is subdivided into time windows 34, which have the same time length and which is, for example, equal to one millisecond. Once the last time window 34 is completed, this time window 34 is divided into a number of frequency channels 36, such as in FIG. 4 shown. For division of the input signal 32 into the individual frequency channels 36, bandpass filters (frequency-pass filters) 38 which are present within the signal processing unit 16 are used. Each of the frequency channels 36 is assigned a particular current channel value 40. In summary, in the second step 30, the input signal 32 is divided into the individual frequency channels 36 and discretized by means of the assignment of the current channel value 40.

Furthermore, after execution of the first working step 16, a third operation 42 is carried out on the transmitter 18 side, in which channel values 44 preceding in time are filled. These have been determined, for example, in a previous run of the method 26 or, if the method 26 has not yet been carried out, a standard value is used for this purpose. Also, a fourth step 46 is performed on the receiver 20 side, in which a reconstructed audio signal 48 preceding the time is determined. This corresponds to the temporally preceding channel values 44 and is determined in the same way as the temporally preceding channel values 44.

Furthermore, a fifth step 50 is carried out in which the number of current channel values 40 are divided into a first current data record 52 and a second current data record 54. Here, the current channel values 40 associated with an odd frequency channel 36 are assigned to the first data set 52 and the remaining current channel values 40 are assigned to the second current data set 54 so that the two current data sets 52, 54 have substantially the same number of current channel values 40 ,

In a sixth step 56, a first prediction 58 for the first current data set 52 is created on the basis of the previously preceding channel values 44. To generate the first forecast 58, a linear prediction is used. In other words, a number of forecast values 60 are created, wherein each of the forecast values 60 is associated with one of the current channel values 40 of the first current data set 52. For example, only the temporally preceding channel values 44 are used, which are assigned to the temporally directly preceding time window 34. Here are to create the respective Forecast value 60 ("precisified value"), for example, uses only the temporally preceding channel values 44 associated with adjacent frequency channels 36.

In a subsequent seventh operation 61, a first deviation 60 is created between the first prediction 58 and the first current data set 52, for which the difference between each of the current channel values 40 of the first current data set 52 and each of the prediction values 60 of the first prediction 62 deducted from a difference value. The number of forecast values 60 and the number of difference values correspond to the number of current channel values 40 of the first current data set 52.

In a subsequent eighth step 64, a first transmission value 66 is generated by means of a spherical logarithmic quantization, which corresponds to the first deviation 62. The first transmission value 66 is one-dimensional. In other words, the multidimensional first deviation 62 is assigned the one-dimensional first transmission value 66. This is transmitted by means of one of the data signals 22 to the receiver 20 of the remaining hearing aid 4. A ninth step 68 is carried out by means of the communication device 24 of the hearing aid 4 having the receiver 20, in which a first auxiliary data set 70 is created by means of the inverse function corresponding to the quantization, which thus corresponds to the first deviation 62. Further, using the temporally preceding reconstructed output signal 48, a first auxiliary prediction 72 is also generated with prediction values, using the same linear prediction as for generating the first prediction 58. Since the temporally preceding reconstructed output signal 48 corresponds to temporally preceding channel values 44 the first auxiliary prediction 72 of the first prognosis 58. Furthermore, the first auxiliary data set 70 is added to the first auxiliary prognosis 72 in a manner that is valid in terms of value. The resulting data set is a first reconstructed data set 74 which, with the exception of any noise / interference induced due to the use of the spherical logarithmic quantization, corresponds to the first data set 52 present on the transmitter 18 side.

On the part of the transmitter 18, using the current channel values 44 of the first current data record 52 by means of a linear prediction in a tenth operation 76, a second prognosis 78 is calculated with a number of current channel values 44 of the second current data record 54 corresponding number of forecast values 80 ("precisified value"). For example, the channel values 44 of adjacent frequency channels 36 of the first current data set 52 and the respective temporally preceding value of the same frequency band 36 will be used to generate the prediction values 80. Consequently, the second prediction 78 is generated based on the temporally preceding channel values 44 and the first current data set 52. Each of the prediction values 80 corresponds to one of the current channel values 44 of the second current data set 54. In a subsequent eleventh operation 82, a second deviation 84 between the second prediction 78 and the second current data set 54 is determined by the difference between each current channel value 40 of the second current record 54 and the respectively associated forecast value 80 of the second forecast 78 is created to form a difference value. The number of difference values forms the second deviation 84.

In a subsequent twelfth step 86, a second transmission value 88 is generated by means of the second deviation 84 by means of spherical logarithmic quantization, which consequently corresponds to the second deviation 84. The second transmission value 86 is also one-dimensional in this case, and, for example, substantially the same spherical logarithmic quantization is used, which is also used to generate the first transmission value 66. The second transmission value 88 is transmitted by the transmitter 18 by means of the data signals 22 to the receiver 20 of the hearing aid 4, to which the first transmission value 66 has also been transmitted.

By means of the communication device 24, a second reconstructed data record 92 is created in a thirteenth operation 90. In this case, based on the second transmission value 88, a second auxiliary data record 94 is created, wherein a for spherical logarithmic quantization inverse function is performed. Consequently, the second auxiliary data set 94, with the exception of any noise introduced as a result of the quantization, corresponds to the second deviation 84. Further, in the thirteenth step 90, the first reconstructed data set 74 and the preceding reconstructed output signal 48 and the first reconstructed one Dataset 74 created a second auxiliary prognosis 96, for which a linear prediction is used. In this case, the same coefficients are used as for the generation of the second prediction 78. Therefore, the second auxiliary prediction 96 substantially corresponds to the second prediction 78 due to the substantially same values used to generate the prediction Werteweise the second auxiliary prognosis 96 added. In summary, the second reconstructed data set 92 is produced on the receiver side based on the temporally preceding reconstructed output signal 48 and the first reconstructed data set 74.

In a subsequent fourteenth operation 98, the first reconstructed data record 74 and the second reconstructed data record 92 are combined into a reconstructed output signal 100, which consequently essentially has the current channel values 40. Any difference is present only due to any quantization effects. The output signal 100 is used in a renewed execution of the method 26 as temporally previously reconstructed output signal 48 or at least added to this. In a subsequent fifteenth operation 102, the reconstructed output signal 100 is transformed from the frequency space to the time period and output, for example, by means of the loudspeaker 14.

Furthermore, a sixteenth operation 104 is carried out on the transmitter 18 side, in which a third reconstructed data set 106 is created on the basis of the first and second transmission values 66, 88 as well as the first and second prognoses 58, 78. In this case, the ninth step 68 and the thirteenth step 90 are carried out essentially on the part of the transmitter 18, the first prognosis 58 instead of the first auxiliary prognosis 72 and the first prognosis second prognosis 78 is used instead of the second auxiliary prognosis 96. Also, the two reconstructed data sets are added to form the third reconstructed data set 106. As a result, the third reconstructed data set 106 corresponds to the output signal 100. In other words, the third reconstructed data set 106 also has any noise due to the quantization used. The third reconstructed data record 106 is used in a renewed execution of the method 26 as temporally preceding channel values 44, so that both on the side of the transmitter 18 and on the side of the receiver 20 for generating the respective forecasts 58, 72, 78, 96, the respective same input data be used.

In FIG. 5 is a further embodiment of the preparation of the first prediction 58 and the creation of the second prognosis 78 and consequently also the creation of the two reconstructed data sets 74, 92 shown. Here, the first current data record 54 has only one of the current channel values 44, and consequently the first forecast 58 only contains a single forecast value 60. Based on the current channel value 40 assigned to this single forecast value 60, one of the forecast values 80 of the second prognosis 78 is created. which is assigned to the directly adjacent frequency channel 36. The current channel value 40 associated with this prognosis value 80 is again used in a further working step to determine a further prognosis value 80 which is assigned to the same in the directly adjacent frequency channel 36, etc. Alternatively, all prognosis values 80 are based on the single current channel value 40 of the first actual value Record 54 created.

In FIG. 6 is shown a further embodiment. In this case, the first data record 52 substantially comprises one fifth of all current channel values 40, wherein these are assigned, for example, to frequency channels 36 which are spaced apart from each other by means of four of the frequency channels 36. For determining the prognosis values 80 of the second prognosis 78 and consequently also of the second auxiliary prognosis 96, these are used, the number of the prognosis value 80 corresponding to the number of prognosis values 60. Following this, by means of the prognostic values 80 of the second prognosis 78, the current ones are assigned Channel values 40 further prediction values 80 which are assigned to the directly adjacent frequency band 36 are determined. For example, the current channel values 40 are divided into a plurality of data records, and a number of deviations are determined and a corresponding transmission value corresponding thereto is transmitted. In other words that will be in FIG. 2 illustrated method 26 executed essentially cascaded.

In summary, due to the use of already reconstructed or actual values for generating the second prognosis 78 or for generating the second auxiliary prognosis 96, a possible correlation between the frequency channels 36 is taken into account so that the second deviation 84 is comparatively small.

The invention is not limited to the embodiments described above.

LIST OF REFERENCE NUMBERS

2

Hearing aid system

4

hearing aid

6

casing

8th

microphone

10

transducer

12

Signal processing unit

14

speaker

16

audio signal

18

transmitter

20

receiver

22

data signal

24

communicator

26

method

28

first step

30

second step

32

input

34

Time window

36

frequency channel

38

Bandpass filter

40

current channel value

42

third step

44

temporally preceding channel value

46

fourth step

48

temporally preceding reconstructed output signal

50

fifth step

52

first current record

54

second current record

56

sixth step

58

first prognosis

60

forecast value

61

seventh step

62

first deviation

64

eighth step

66

first transmission value

68

ninth step

70

first auxiliary data set

72

first auxiliary prognosis

74

first reconstructed record

76

Tenth work step

78

second prognosis

80

forecast value

82

eleventh step

84

second deviation

86

Twelfth step

88

second transmission value

90

thirteenth work step

92

second reconstructed record

94

second auxiliary data set

96

second auxiliary prognosis

98

Fourteenth step

100

reconstructed output signal

102

fifteenth step

104

sixteenth step

106

third reconstructed record

Claims (8)

1. Method (26) for transmitting an audio signal (16) from a transmitter (18) to a receiver (20), in which

   - at the transmitter end, an input signal (32) corresponding to the audio signal (16) is divided into a number of frequency channels (36) for a particular time window (34),

   - at the transmitter end, a current channel value (40) is allocated to each frequency channel (36), the current channel value (40) being an amplitude and/or a phase or a signal level,

   - at the transmitter end, the current channel values (40) are divided into a first current data record (52) and a second current data record (54), the first current data record (52) and the second current data record (54) in each case comprising at least one of the current channel values (40),

   - at the transmitter end, a first forecast (58) is generated for the first current data record (52) by means of a channel value (44) preceding in time, a linear prediction being utilized for generating the first forecast (58),

   - at the transmitter end, a first deviation (62) between the first forecast (58) and the first current data record (52) is determined, the first deviation (62) being generated in that the difference between each current channel value (40) of the first current data record (52) and an allocated prognostic value (60) of the first forecast (58) is generated for forming a difference value and the difference values form the first deviation (62),

   - at the transmitter end, a second forecast (78) is generated for the second current data record (54) by means of the first current data record (52), a linear prediction being utilized for generating the second forecast (78),

   - at the transmitter end, a second deviation (84) between the second forecast (78) and the second current data record (54) is determined, the second deviation (84) being generated in that the difference between each current channel value (40) of the second current data record (5) and an allocated prognostic value (80) of the second forecast (78) is generated for forming a difference value and the difference values form the second deviation (84),

   - a first transmission value (66) corresponding to the first deviation (62) is transmitted from the transmitter (18) to the receiver (20),

   - a second transmission value (88) corresponding to the second deviation (84) is transmitted from the transmitter (18) to the receiver (20),

   - at the receiver end, a first reconstructed data record (74) is generated by means of a reconstructed output signal (48) preceding in time and the transmitted first transmission value (66), a linear prediction being utilized for generating the first reconstructed data record (74),

   - at the receiver end, a second reconstructed data record (92) is generated by means of the first reconstructed data record (74) and the transmitted second transmission value (88), a linear prediction being utilized for generating the second reconstructed data record (92) and,

   - at the receiver end, the first reconstructed data record (74) and the second reconstructed data record (92) are combined to form a reconstructed output signal (100) and at the transmitter end, the second forecast (78) being also generated by means of the channel value (44) preceding in time, and at the receiver end, the second reconstructed data record (92) being also generated by means of the reconstructed output signal (48) preceding in time.
2. Method (26) according to Claim 1,
   characterized in that
   the input signal (32) is divided into the frequency channels (36) by means of bandpass filters (38).
3. Method (26) according to one of Claims 1 to 2,
   characterized in that
   the first transmission value (66) is generated by means of quantization of the first deviation (64) and/or that the second transmission value (84) is generated by means of quantization of the second deviation (84).
4. Method (26) according to Claim 3,
   characterized in that
   a vector quantization is utilized for the quantization.
5. Method (26) according to one of Claims 3 to 4,
   characterized in that
   a spherical logarithmic quantization is utilized for the quantization.
6. Method (26) according to one of Claims 1 to 5,
   characterized in that
   the current channel values (40) are essentially divided in halves between the first current data record (52) and the second current data record (54) .
7. Hearing device (4), particularly hearing aid, comprising a communication facility (24) which is provided and configured for transmitting and/or receiving an audio signal (16) according to a method (26) according to one of Claims 1 to 6.
8. Hearing device system (2) having two hearing devices (4) according to Claim 7, which is provided and configured to transmit audio signals (16) between the two hearing devices (4) by means of their communication facilities (24) according to a method (26) according to one of Claims 1 to 6.

Images