WO2019227353A1

WO2019227353A1 - Method and device for estimating a direction of arrival

Info

Publication number: WO2019227353A1
Application number: PCT/CN2018/089067
Authority: WO
Inventors: Andras Palfi; Areeb RIAZ
Original assignee: Goertek Inc.
Priority date: 2018-05-30
Filing date: 2018-05-30
Publication date: 2019-12-05
Also published as: CN108713323A; CN108713323B

Abstract

This disclosure relates to a method and device for estimating a direction of arrival. The method comprises: collecting signals from two or more sensors; obtaining a first direction of arrival according to at least one part of the collected signals when it is determined that the signals include voice and noise; obtaining a stored second direction of arrival for noise; obtaining an SNR used for present estimation according to at least one part of the signals; and modifying the first direction of arrival to remove its bias caused by the noise based on the second direction of arrival and the present SNR to obtain a de-biased direction of arrival.

Description

METHOD AND DEVICE FOR ESTIMATING A DIRECTION OF ARRIVAL

FIELD OF THE INVENTION

The present invention relates to microphone signal processing, and more specifically, to a method and device for estimating a direction of arrival.

BACKGROUND OF THE INVENTION

In the technical field of audio signal processing for a microphone array, by exploiting spatial characteristics of a sound field, it is possible to process the incoming sound signals so that only sound waves incident from a certain angle or range of angles are retained, while sound waves from other directions are filtered away.

One approach to achieving such spatial filtering is to use a family of algorithms called beam-forming.

In many applications of processing signals from a sensor array, the angle at which the sound source of interest (termed “desired source” or “valuable source” ) is located relative to the array is not known a priori. In these applications, one of the main challenges is to be able to determine the angle of the valuable source from the audio data (or signals) itself. It is a problem that is termed DOA (direction of arrival) estimation. This is often accomplished by utilizing several beam-formers pointing towards different directions, and analyzing their outputs in some way. This is, however, a computationally expensive approach.

Another approach for DOA estimation relies on the fact that a sound wave’s incident angle at a microphone array is directly related to the phase differences observed between the sound signals captured by a pair (or pairs) of microphones in the array. This approach is computationally less expensive than the beam-forming approach. For a given microphone array geometry and a given assumption on the shape of the incident sound wave-front (which in the majority of cases –although not exclusively -is assumed to be a plane wave) , an analytical formula can be derived for the relation between the angle of the sound source location and the observed phase difference between the microphones. Thus, by analyzing the phase information in the captured audio data and by utilizing the above-mentioned relation between incident angle and phase difference, the angle at which a sound source is located can be inferred.

The challenge in estimating the DOA of the desired source from the observed phase differences is that apart from the valuable sound signal, the acoustical background noise and the electrical measurement noise also contribute to the observed phases. Estimating the DOA of the desired source based on these phase measurements might, therefore, yield wrong results, especially if the contribution from the noise sources is too large.

SUMMARY OF THE INVENTION

One object of this invention is to provide a new technical solution for estimating a direction of arrival.

According to a first aspect of the present invention, there is provided a method for estimating a direction of arrival, comprising: collecting signals from two or more sensors; obtaining a first direction of arrival according to at least one part of the collected signals when it is determined that the signals include valuable signal and noise; obtaining a stored second direction of arrival for noise; obtaining an SNR used for present estimation according to at least one part of the signals; and modifying the first direction of arrival to remove its bias caused by the noise based on the second direction of arrival and the present SNR to obtain a de-biased direction of arrival.

According to a second aspect of the present invention, there is provided a device for estimating a direction of arrival, comprising: two or more sensors, which collect signals; an SNR module, which obtains an SNR used for present estimation according to at least one part of the signals; a DOA module, which obtains a first direction of arrival according to at least one part of the collected signals when it is determined that the signals include valuable signal and noise; and a de-biased DOA module, which obtains a stored second direction of arrival for noise and modifies the first direction of arrival to remove its bias caused by the noise based on the second direction of arrival and the estimated SNR to obtain a de-biased direction of arrival.

According to an embodiment of this disclosure, the present invention can improve the accuracy for estimation of direction of arrival.

Further features of the present invention and advantages thereof will become apparent from the following detailed description of exemplary embodiments according to the present invention with reference to the attached drawings.

BRIEF DESCRIPTION OF THE DRAWINGS

The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate embodiments of the invention and, together with the description thereof, serve to explain the principles of the invention.

Fig. 1 is a graph showing a relationship between phase differences and frequency captured by two microphones in an idealized scenario.

Fig. 2 is a graph showing a relationship between phase differences and frequency captured by two microphones in a non-idealized scenario.

Fig. 3 is a graph showing a best fit line for a relationship between phase differences and frequency captured by two microphones in a non-idealized scenario.

Fig. 4 shows a flowchart of a method for estimating a direction of arrival according to an embodiment.

Fig. 5 shows a diagram of a device for estimating a direction of arrival according to another embodiment.

DETAILED DESCRIPTION OF THE EMBODIMENTS

Various exemplary embodiments of the present invention will now be described in detail with reference to the drawings. It should be noted that the relative arrangement of the components and steps, the numerical expressions, and numerical values set forth in these embodiments do not limit the scope of the present invention unless it is specifically stated otherwise.

The following description of at least one exemplary embodiment is merely illustrative in nature and is in no way intended to limit the invention, its application, or uses.

Techniques, methods and apparatus as known by one of ordinary skill in the relevant art may not be discussed in detail but are intended to be part of the specification where appropriate.

In all of the examples illustrated and discussed herein, any specific values should be interpreted to be illustrative only and non-limiting. Thus, other examples of the exemplary embodiments could have different values.

Notice that similar reference numerals and letters refer to similar items in the following figures, and thus once an item is defined in one figure, it is possible that it need not be further discussed for following figures.

It shall be appreciated by a person skilled in the art that some specific examples are described here in an illustrative manner and are in no way to limit the scope of the invention. Under the teaching of this description, a person skilled in the art can conceive other examples or embodiments which shall also be covered by the claims.

Estimating an angle at which a sound source is located relative to a microphone array can be done for example by analyzing the differences in observed phase between a set of microphones in a microphone array. Fig. 1 shows an ideal simple example of how the phase can relate to the angle of incidence for a pair of microphones in any given microphone array. In Fig. 1, y-axis represents phase difference derived from sound signals captured by two microphones, and x-axis represents the frequency of the sound signals. The dots in Fig. 1 represent the relationship of phase differences and frequencies. It can be seen from Fig. 1 that in an idealized scenario, the dots therein are presented in a line and the slope of the line directly relates to the direction of the incident sound wave.

The scenario depicted in Fig. 1 is highly idealized. In the idealized scenario, the sound source is assumed to be broadband and contain energy at many frequencies, the incident sound wave is assumed to be perfectly planar (which is equivalent to assuming that the sound source is at an infinite distance away from the microphones) , and no background noise is assumed to be present.

However, in reality, this is impossible. Any deviation from these idealized assumptions will disturb the phase differences observed by two microphones. Fig. 2 depicts such a non-idealized scenario. In Fig. 2, the dots are scattered and the slope of the phase difference curve is not perfectly evident anymore, due to the above-mentioned disturbances. The slope of the phase differences is not apparent, because the sound wave captured by a microphone is a combination of many plane waves.

In a scenario like that shown in Fig. 2, the DOA may be estimated by taking the slope of a line that is a “best fit” to the phase difference dots. The best-fit line can, for example, be chosen by the so-called least mean square error rule. Any other fitting algorithm is also possible and may produce an equally good result. Fig. 3 depicts such a scenario of a best-fit line A (dash line) , which is a straight line and is being fit to the phase difference dots scattered around it. The slope of this best-fit line A will be used to estimate the DOA.

In general, a sound field observed by a microphone array is composed of a valuable (desired) sound signal the angle of which we would like to estimate, background noise which is sounds in the environment and the angle of which we would like to filter out and measurement noise which is electrical noise inherent to the microphone array or measurement hardware and is not related to any physical sound angle of arrival. These three components all contribute to the phase information observed in the audio channels. In effect, the contributions from the noises bias the DOA result. In the prior art DOA estimation method, this bias is embodied by the fact that the best-fit line was fitted to a set of data points (such as the dots in Fig. 3) that contain contributions from the valuable source as well as the noise components.

Because the component related to the measurement noise is relatively low and can be controlled via hardware improvement to some degree, it is deemed to be of a negligible amount and thus is not considered here.

It shall be understood by a person skilled in the art that although DOA can be estimated through a microphone array, other kind of sensor array is also possible as long as it can capture direction information. For example, two accelerometers are arranged in an array and the sequence of signals captured by the two accelerometers may render the direction of a source. Other sensors may produce similar results.

As shown in Fig. 4, at step S1100, signals are collected from two or more sensors.

At step S1200, a first direction of arrival is obtained according to at least one part of the collected signals when it is determined that the signals include valuable signal and noise. For example, the valuable signal may be voice signal.

At step S1300, a stored second direction of arrival for noise is obtained.

At step S1400, an SNR used for present estimation is obtained according to at least one part of the signals.

At step S1500, the first direction of arrival is modified to remove its bias caused by the noise based on the second direction of arrival and the present SNR to obtain a de-biased direction of arrival.

The first direction of arrival is the angle at which the desired source is estimated to be located, wherein this estimation of the desired source is biased by noise. The second direction of arrival is the angle at which the noise source is estimated to be located. The first direction arrival is de-biased to remove the bias caused by the second direction of noise to approach a true direction of arrival.

In this method, the direction of arrival is calibrated to remove the bias caused by noise. Because it is difficult to estimate the noise in a current signal which includes voice and noise, a stored direction of arrival for noise is used in the calibration. Furthermore, in removing the bias, the present SNR is used to determine how much noise shall be removed. This will improve the performance of removing noise bias. The direction of arrival may be the angle of arrival for the impinging sound wave.

As explained above, although other sensors can similarly be used to obtain a direction of arrival, the sensors here are preferably microphones in a microphone array, the microphone array includes at least two microphones, and sound signals from microphones of the microphone array are collected.

Under the teaching of this disclosure, many approaches for obtaining the second direction of arrival are possible. For example, it may be obtained from the noise components in the signals previously collected by the sensors.

In this embodiment, rather than just estimating the level of noise, both the level and the location of the surrounding acoustic background noise are estimated and are used to make correct (i.e. de-bias) the estimated DOA (the first direction of arrival) .

Optionally, the second direction of arrival may be obtained according to at least one part of the signals when it is determined that the signals only include noise. Then, the second direction of arrival of the noise is stored and the estimation (i.e. the obtaining of the second direction) of the present time is stopped. This approach may improve the power efficiency of an electric apparatus because the estimation operation of noise will be stopped for repetitive similar noises and will not be running all the time. A designer can set the timing of the estimation operation for the second direction of arrival, for example, periodically or when a change in noise level is detected. The stored second direction of arrival will be used later for calibrating the first direction of arrival. Here, the initial value of the second direction of arrival may be set to a pre-defined value such as “0” .

This approach is simple and efficient because the calculation of the first direction of arrival and the calculation of the second direction of arrival can share the same implementation hardware and/or software. Furthermore, because this approach also uses an SNR for the present estimation of the first direction of arrival, it is also effective. Here, the level of the background noise can be represented by the SNR, and the direction of the background noise can be estimated as the noise-DOA (the second direction of arrival) , which is estimated during inactive periods, i.e. when it is determined that the signals only include noise.

Under the teaching of this disclosure, many approaches for determining that the signals only contains noise. For example, a user can presses a hold button on an electronics device when he speaks. It is determined that the signals only contains noise when a release of the hold button is detected.

Optionally, the determination can be performed by voice activity detection. A voice activity detection is performed on at least part of the collected signals to determine whether the signals include voice. If it is determined that the signals include voice, it is determined that a sound source corresponding to the signals is active. If it is determined that the signals do not include voice, it is determined that the sound source corresponding to the signals is inactive. Here, the second direction of arrival is obtained when the voice activity detection determines that the sound source is inactive, and the first direction of arrival is obtained when the voice activity detection determines that the sound source is active.

For example, the sensors may include microphones in a microphone array or/and an accelerometer, and the signals may be collected from the microphone array or/and the accelerometer. The voice activity detection may be performed based on the signals collected by the microphone array or by the accelerometer, or both of them.

The de-biased direction of arrival can be expressed as DOA _de -biased, and may be obtained through the following relationship:

DOA _de-biased=f (DOA _estimated, SNR _estimated, DOA _noise)

wherein DOA _noise is the second direction of arrival, DOA _estimated is the first direction of arrival, SNR _estimated is the present SNR, and f () is a function for modifying the first direction.

For example, f () is a linear operation of DOA _noise and a difference between DOA _estimated multiplied with a coefficient and DOA _noise multiplied with a coefficient, and the coefficient is a function of the present SNR. In an example, the coefficient multiplied with DOA _estimated is the same as the coefficient multiplied with the DOA _noise.

The coefficient could be any suitable function selected by a technician according to his experiences or experiments. The output of the function of the present SNR should increase, as the noise level increases. However, SNR is inversely proportional to the noise level, i.e. if the noise level increases, the SNR decreases. Therefore, if we want the output of the function of the present SNR to be higher when the noise is higher, the coefficient shall be a custom-defined function with monotonically increasing return values as SNR _estimated decreases. In this manner, it may compensate the loss caused by the noise to a better degree.

For example, the coefficient may be defined in a lookup table which is stored in memory.

In this embodiment, the direction of arrival is not de-biased just based on the value of noise or that of SNR. It is de-biased by the combination of the stored direction of arrival for noise and the current SNR. In this manner, the de-biasing of the first DOA will be based on the level and position of the noise rather than just the value of the noise, and the amount of de-bias applied to the first DOA will depend on the amount of determined SNR. The SNR may be estimated from the current signals and will be used to change the amount of de-biasing. For example, the DOAs may be represented as angles of arrival, and vector operations for the angles of arrival may be used to remove the bias caused by the noise DOA.

In this embodiment, a more accurate estimation of where a sound source is located is obtained to de-bias the DOA so that "distortions" added by the external noise are removed.

The processing before the estimations of DOA such as low-pass filtering can be those known in the art and thus the detailed description thereof is omitted.

Generally, the background noise component adds a varying amount of bias to a DOA estimation result. The inventor’s experience shows that the bias is dependent on the level of the noise and the noise DOA. The noise DOA relates to the direction towards which the estimated DOA is biased, and the level of the noise relates to the amount of bias added to the estimated DOA. Here, the actual location of the noise source does not matter, but the angle at which it is estimated to be located matters. For example, for diffusing noises which are arriving from many directions, the DOA estimation method would estimate a noise DOA of around zero degrees. It shall be appreciated by a person skilled in that depending on the setup, the DOA algorithm used and the definition of coordinate-system, the estimated noise angle for diffusing noises might also be some other value.

In this embodiment, a stored DOA for noise is used. It may be obtained in the previous processing of DOA, for example, when the desired sound source is inactive. This is effective and efficient when it is used together with the current SNR. The initial value for the store DOA may be set to be zero degree or any other degree chosen a designer when the electronics apparatus is powered on and the store DOA can be updated during the inactive period.

The calculation for the voice DOA and the calculation for the noise can share the same hardware and/or software components. The output of the components can be stored as a noise DOA when the sound source is inactive or as an estimated DOA for the sound source when it is active.

The device 500 shown in Fig. 5 could be a part of an electronics apparatus, which requires information of direction of arrival. For example, it may use this information to perform a noise cancellation. The device 500 in Fig. 5 can be used to implement the above method and thus the repetitive description will be omitted.

As shown in Fig. 5, the device 500 for estimating a direction of arrival comprises: two or more sensors 101-1, 101-2, …101-n; a DOA module 201; an SNR module 202; and a de-biased DOA module 301.

The two or more sensors 101-1, 101-2, …101-n collect signals.

The SNR module 202 obtains an SNR (signal-to-noise ratio) used for present estimation according to at least one part of the signals. As shown in Fig. 5, the SNR module 202 takes, as its input, a signal or a set of signals from a microphone array and provides, as its output, an estimation of the SNR.

The DOA module 201 obtains a first direction of arrival according to at least one part of the collected signals when it is determined that the signals include voice and noise. Optionally, the DOA module 201 can use the algorithms known in the prior art. This disclosure is not directed a specific algorithm of DOA and thus is not to it. It receives signals from sensors such as microphones and outputs information indicating an angle at which a sound source is estimated to be located.

The de-biased DOA module 301 obtains a stored second direction of arrival for noise and modifies the first direction of arrival to remove its bias caused by the noise based on the second direction of arrival and the estimated SNR to obtain a de-biased direction of arrival.

The device 500 in this embodiment further comprises a memory (not shown) . The memory stores the stored second estimated direction of arrival for noise. The DOA module 201 obtains a second direction of arrival according to at least one part of the signals when it is determined that the signals only include noise and stores the second direction of arrival in the memory as the stored second estimated direction of arrival for noise. The de-biased DOA module 301 obtains the second estimated direction of arrival from the memory to modify the first direction of arrival.

Alternatively, the device 500 may further comprise a VAD (Voice Activity Detector) module 203. The VAD module 203 performs a voice activity detection to determine whether the signals include voice. If the VAD module 203 determines that the signals include voice, it determines that a sound source corresponding to the signals is active. If the VAD module 203 determines that the signals do not include voice, it determines that the sound source corresponding to the signals is inactive. The DOA module 201 obtains the second estimated direction of arrival when the VAD module 203 determines that the desired sound source is inactive and obtains the first direction of arrival when the VAD module 203 determines that the desired sound source is active.

The VAD module 203 may perform the voice activity detection based a signal from an accelerometer.

The VAD module 203 can detect whether the desired sound source is active in a given signal segment. It can takes, as its input, a signal or a set of signals from the microphone array or optionally other sensor such an accelerometer and provides, as its output, a true or false decision based on whether contribution from the desired sound source is detected in the input signal or signals or whether the desired sound source is producing a sound.

The sensors 101-1, 101-2, …101-n shown in Fig. 5 may include microphones in a microphone array or/and an accelerometer. The microphone array or/and the accelerometer collect the signals.

Although Fig. 5 shows that the outputs of the two or more sensors 101-1, 101-2, …101-n are connected to the inputs of the DOA module 201, the SNR module 202 and the VAD module 203, respectively, these modules can just receive some of them. For example, the sensors include microphones in a microphone array and an accelerometer. In this case, the DOA module 201 receives the signals from at least two of the microphones, the SNR module 202 receives signals two or more microphones, and the VAD module 203 receives signals from the accelerometer.

In an example, the de-biased DOA module 301 obtains the de-biased direction of arrival DOA _de-biased through the following relationship:

DOA _de-biased=f (DOA _estimated, SNR _estimated, DOA _noise)

wherein DOA _noise is the second estimated direction of arrival for noise, DOA _estimated is the first direction of arrival for the desired sound source, SNR _estimated is the estimated SNR, and f () is a function for modifying the first direction of arrival. f () may be a linear operation of DOA _noise and a difference between DOA _estimated multiplied with a coefficient and DOA _noise multiplied with the coefficient, and the coefficient is a function of the present SNR.

Preferably, the coefficient is a custom-defined function with monotonically increasing return values as SNR _estimated decreases.

The sensors 101-1, 101-2, …101-n shown in Fig. 5 may be microphones in a microphone array, and the microphone array includes at least two microphones. The at least two microphones collect sound signals.

For example, at least one of the SNR module, the DOA module, the VAD module and the de-biased DOA module is implemented in at least one of a discrete device, a DSP, a programmable device, an ASIC and a combination of a processor and a memory.

It shall be understood by a person skilled in the art that this disclosure is not directed to the improvements of the SNR module, the DOA module and the VAD module per se and they can adopt prior art algorithms to implement their functions. So, the detailed description of them are omitted.

It will be understood by a person skilled in the prior art that a software is equivalent to a hardware except for some of the mechanical components such a speaker, a microphone and so on. In this regard, a person skilled in the art can conceive, under the teaching of this disclosure, that the processing of any of the DOA module 201, the SNR module 202, the VAD module 203 and the de-biased DOA module 301 can be carried out through a hardware manner, a software manner and/or a combination thereof. For example, it can be carried out through discrete devices, ASIC, a programmable device such PLD, DSP, FPGA. Alternatively, it can be implemented in a combination of a computing device such as a CPU or a MPU and a memory, wherein instructions are stored in the memory and are used to control the computing device to performing corresponding operations during the running of the earphone. In this regard, this disclosure will not limit the implementation manners of them. A person skilled in the art can choose the implementation manners under the teaching of this disclosure in consideration of the cost, the market availability and so on.

Although some specific embodiments of the present invention have been demonstrated in detail with examples, it should be understood by a person skilled in the art that the above examples are only intended to be illustrative but not to limit the scope of the present invention.

Claims

A method for estimating a direction of arrival, comprising:

collecting signals from two or more sensors;

obtaining a first direction of arrival according to at least one part of the collected signals when it is determined that the signals include voice and noise;

obtaining a stored second direction of arrival for noise;

obtaining an SNR used for present estimation according to at least one part of the signals; and

modifying the first direction of arrival to remove its bias caused by the noise based on the second direction of arrival and the present SNR to obtain a de-biased direction of arrival.
The method according to claim 1, further comprising:

obtaining a second direction of arrival according to at least one part of the signals when determining that the signals only include noise;

storing the second direction of arrival of the noise; and

stopping the obtaining of the second direction of arrival of the present time.
The method according to claim 1, further comprising:

performing a voice activity detection on at least part of the collected signals to determine whether the signals include voice,

wherein if it is determined that the signals include voice, it is determined that a sound source corresponding to the signals is active; and if it is determined that the signals do not include voice, it is determined that the sound source corresponding to the signals is inactive,

wherein the second direction of arrival is obtained when the voice activity detection determines that the sound source is inactive;

the first direction of arrival is obtained when the voice activity detection determines that the sound source is active.
The method according to claim 1 or 3, wherein the sensors include microphones in a microphone array or/and an accelerometer, and collecting signals from two or more sensors includes:

collecting signals from the microphone array or/and the accelerometer.
The method according to claim 1 or 2, wherein the de-biased direction of arrival DOA _de-biased is obtained through the following relationship:

DOA _de-biased=f (DOA _estimated, SNR _estimated, DOA _noise)

wherein DOA _noise is the second direction of arrival, DOA _estimated is the first direction of arrival, SNR _estimated is the present SNR, and f () is a function for modifying the first direction;

wherein f () is a linear operation of DOA _noise and a difference between DOA _estimated multiplied with a coefficient and DOA _noise multiplied with the coefficient, and the coefficient is a function of the present SNR.
The method according to claim 5, wherein the coefficient is a custom-defined function with monotonically increasing return values as SNR _estimated decreases.
The method according to claim 1, wherein the sensors are microphones in a microphone array, and the microphone array includes at least two microphones,

wherein collecting signals from two or more sensors comprising: collecting sound signals from microphones of the microphone array.
A device for estimating a direction of arrival, comprising:

two or more sensors, which collect signals;

an SNR module, which obtains an SNR used for present estimation according to at least one part of the signals;

a DOA module, which obtains a first direction of arrival according to at least one part of the collected signals when it is determined that the signals include voice and noise; and

a de-biased DOA module, which obtains a stored second direction of arrival for noise and modifies the first direction of arrival to remove its bias caused by the noise based on the second direction of arrival and the estimated SNR to obtain a de-biased direction of arrival.
The device according to claim 8, further comprising:

a memory, which stores the stored second estimated direction of arrival for noise,

wherein the DOA module obtains a second direction of arrival according to at least one part of the signals when it is determined that the signals only include noise and stores the second direction of arrival in the memory as the stored second estimated direction of arrival for noise; and

wherein the de-biased DOA module obtains the second estimated direction of arrival from the memory to modify the first direction of arrival.
The device according to claim 8 or 9, further comprising:

a VAD module, which performs a voice activity detection to determine whether the signals include voice,

wherein if the VAD module determines that the signals include voice, it determines that a sound source corresponding to the signals is active; and if the VAD module determines that the signals do not include voice, it determines that the sound source corresponding to the signals is inactive,

wherein the DOA module obtains the second estimated direction of arrival when the VAD module determines that the desired sound source is inactive and obtains the first direction of arrival when the VAD module determines that the desired sound source is active.
The device according to claim 10, wherein the VAD module performs the voice activity detection based a signal from an accelerometer.
The device according to claim 8 or 9, wherein the sensors include microphones in a microphone array or/and an accelerometer, and the microphone array or/and the accelerometer collect the signals.
The device according to claim 8 or 9, wherein the de-biased DOA module obtains the de-biased direction of arrival DOA _de-biased through the following relationship:

DOA _de-biased=f (DOA _estimated, SNR _estimated, DOA _noise)

wherein DOA _noise is the second estimated direction of arrival for noise, DOA _estimated is the first direction of arrival for the desired sound source, SNR _estimated is the estimated SNR, and f () is a function for modifying the first direction of arrival;

wherein f () is a linear operation of DOA _noise and a difference between DOA _estimated multiplied with a coefficient and DOA _noise multiplied with the coefficient, and the coefficient is a function of the present SNR..
The device according to claim 13, wherein the coefficient is a custom-defined function with monotonically increasing return values as SNR _estimated decreases.
The device according to claim 8, wherein the sensors are microphones in a microphone array, and the microphone array includes at least two microphones,

wherein the at least two microphones collect sound signals.
The device according to claim 8 or 9, wherein at least one of the SNR module, the DOA module and the de-biased DOA module is implemented in at least one of a discrete device, a DSP, a programmable device, an ASIC and a combination of a processor and a memory.