JP2022547493A - Wind noise detection - Google Patents

Abstract

【wrap up】
Aspects of the disclosed embodiments provide new methods for detecting wind or handling noise in devices such as headsets or mobile communication devices. The presence of wind noise is determined by calculating the ratio between the microphone signal power and the beamforming signal power and comparing the ratio to a threshold. One value can identify the presence of wind noise and the other value can identify the absence of wind noise. The methods of the disclosed embodiments are also computationally efficient and easy to combine with other processing in devices such as mobile communication devices.

Description

Aspects of this disclosure relate generally to wind noise detection, and more particularly to wind noise detection in devices such as headsets that use multiple microphones.

When using a headset or any mobile device outdoors in windy conditions, wind noise can corrupt the microphone signal. It can also induce wind noise if the user is moving fast, eg when running, walking or biking. Wind noise can negatively affect the phone call, hearing, or enhanced hearing capabilities of these types of devices.

Beamforming may be used in devices with multiple microphones to increase the signal-to-noise ratio (SNR). However, with beamformed signals, the wind noise is even greater. There are certain phase and amplitude differences between different microphone signals, depending on the direction in which they arrive at the microphone array. During wind noise, the phase and amplitude differences of the microphone signals do not have the assumed relationship. In a worst-case scenario, the microphone signals can add during the wind noise, resulting in twice the wind noise in the beamformed signal compared to the microphone signal. Also, handling noise, such as when a user touches a microphone or microphone holes, can add up in a similar manner as wind noise.

There are several ways to detect wind noise. For example, high power at low frequencies can indicate wind noise. The microphone signals can be subtracted from each other and a large difference can indicate wind noise. Additionally, wind noise can be detected by calculating the ratio of the difference and the sum of the microphone signals and comparing the result to a threshold. However, these countermeasures are prone to errors and lead to false positives.

Accordingly, it would be desirable to be able to detect wind noise in a manner that addresses at least some of the problems identified above.

It is an object of the disclosed embodiments to provide wind detection in devices such as headsets or mobile communication devices. This object is solved by the subject matter of the independent claims. Further advantageous variants can be found in the dependent claims.

According to a first aspect, the above and additional objects and advantages are accomplished by an apparatus that includes a processor. In one embodiment, the processor determines a microphone signal power of at least one microphone of the device; monitors beamforming signal power of at least two microphones of the device; and compares the microphone signal power to the beamforming signal power. configured to detect wind noise in the device's microphone signal based on the comparison; Aspects of the disclosed embodiment reliably detect wind noise and reduce false positives. This method is computationally efficient and easy to combine with other processing in devices such as mobile communication devices or headsets.

In a possible embodiment of the apparatus, the processor compares the microphone signal power to the beamformed signal power by calculating a ratio of the microphone signal power to the beamformed signal power; It is further configured to detect noise. Aspects of the disclosed embodiment reliably detect wind noise from the power ratio between the microphone signal power and the beamforming signal power. This process is computationally efficient.

In a possible embodiment of the device, the processor is further arranged to detect wind noise if the calculated ratio is less than a predetermined threshold. Aspects of the disclosed embodiment reliably detect wind noise from the power ratio between the microphone signal power and the beamforming signal power. This ratio may be 1 for sound coming from the target direction, greater than 1 for ambient noise, and less than 1 for wind noise.

In a possible embodiment of the device, the processor switches off beamforming of the at least two microphones when wind noise is detected; is further configured to select the microphone for which an amount of wind noise is detected. Once wind noise is detected, beamforming can be switched off and wind noise can be reduced using only single microphone processing. The microphone with the lowest amount of wind noise is selected.

In a possible embodiment of the device, the processor switches off beamforming of the at least two microphones when wind noise is detected; It is further configured to select. Once wind noise is detected, beamforming can be switched off and wind noise interference can be reduced using only single microphone processing. One microphone may be selected, for example the one closer to the mouth than the other.

In a possible embodiment of the apparatus, the processor is further configured to compare the other microphone signal power of at least one of the at least two microphones with the beamforming signal power and detect wind noise based on the comparison. It is Detection is enhanced by monitoring more than one microphone of the device. There may be more than two microphones and there is no upper limit on the number of microphones.

In a possible embodiment of the apparatus, the processor is further arranged to calculate the ratio of the microphone signal power and the beamforming signal power over the frequency band. Wind detection is enhanced by calculating the ratio of microphone signal power to beam power over a frequency band.

In a possible embodiment of the apparatus, the processor calculates a ratio of microphone signal power to beamforming signal power over a plurality of frequency bands, and calculates the calculated ratio over a plurality of frequency bands for comparison with a predetermined threshold. It is further configured to select a minimum value of power. Wind noise detection is enhanced by calculating the signal-to-beam power ratio over frequency bands and selecting the minimum value over these bands.

In a possible embodiment of the device, the processor is further arranged to average the calculated power ratios over time. Also, wind noise detection can be enhanced by averaging the power ratio or the minimum value of the power ratio over time over a frequency band.

In a possible embodiment of the device, the processor is further arranged to determine the ambient noise level N. The processor is configured to bypass wind noise detection if the ambient noise level is less than a predetermined threshold.

In a possible embodiment of the device, the device comprises at least two microphones.

In a possible embodiment of the device, the device is a mobile communication device.

In a possible embodiment of the device, the device is an audio signal processing device.

In a possible embodiment of the device, the device is a headset including at least two microphones.

In a possible embodiment of the device, the device is a smartwatch containing at least two microphones.

In a possible embodiment of the device the device is a wearable comprising at least two microphones.

According to a second aspect, the above and additional objects and advantages are achieved by a method of detecting wind noise in a microphone signal of an apparatus. In one embodiment, determining the power of the microphone signal of at least one microphone of the device; determining the power of the beamforming signal of at least two microphones of the device; and comparing the microphone signal power to the beamforming signal power. and detecting wind noise in the device's microphone signal based on the comparison. Aspects of the disclosed embodiment reliably detect wind noise and reduce false positives. This method is computationally efficient and easy to combine with other processing in devices such as mobile communication devices or headsets.

In a possible embodiment of the method, the method comprises the steps of comparing the microphone signal power to the beamformed signal power by calculating a ratio of the microphone signal power to the beamformed signal power; and detecting wind noise in the microphone signal. Aspects of the disclosed embodiment reliably detect wind noise from the power ratio between the microphone signal power and the beamforming signal power. This process is computationally efficient.

In a possible embodiment of the method, the method further comprises detecting wind noise if the calculated ratio is below a predetermined threshold. Wind noise is detected when the signal-to-beam power ratio is less than a lower threshold, and detection is cleared only when the signal-to-beam power ratio is greater than a higher threshold. Threshold settings may differ depending on the product and on characteristics such as microphone position, beam design, and other parameter values.

In a possible embodiment of the method, the method comprises switching off beamforming of at least two microphones when wind noise is detected; and selecting the microphone for which the least amount of wind noise is detected. Once wind noise is detected, beamforming can be switched off and wind noise can be reduced using only single microphone processing. The microphone with the lowest amount of wind noise is selected.

In a possible embodiment of the method, the method comprises switching off beamforming of at least two microphones when wind noise is detected; and selecting a microphone. Once wind noise is detected, beamforming can be switched off and wind noise can be reduced using only single microphone processing. One microphone may be selected, for example the one closer to the mouth than the other.

In a possible embodiment of the method, the method comprises the steps of comparing the beamforming signal power with the microphone signal power of another one of the at least one microphone of the device, and detecting wind noise based on the comparison. and further including Detection is enhanced by monitoring more than one microphone of the device. There may be more than two microphones and there is no upper limit on the number of microphones.

In a possible embodiment of the method, the comparing step further comprises calculating a ratio of microphone signal power to beamforming signal power over the frequency band. Wind detection is enhanced by calculating the ratio of microphone signal power to beam power over a frequency band.

In a possible embodiment of the method, the comparing step includes: calculating a ratio of microphone signal power to beamforming signal power over a plurality of frequency bands; selecting; and comparing the selected minimum value to a predetermined threshold to determine wind noise. Wind noise detection is enhanced by calculating the signal-to-beam power ratio over frequency bands and selecting the minimum value over these bands.

In a possible embodiment of the method, the calculated power ratios are averaged over time. Also, wind noise detection can be enhanced by averaging the power ratio or the minimum value of the power ratio over time over a frequency band.

In a possible embodiment of the method, the method comprises determining the ambient noise level N. Wind noise detection is bypassed if the ambient noise level is below a predetermined threshold.

In a possible embodiment of the method the device comprises at least two microphones.

In a possible embodiment of the method the apparatus is a mobile communication device.

In a possible embodiment of the method the apparatus is an audio signal processing device.

In a possible embodiment of the method the device is a headset comprising at least two microphones.

In a possible embodiment of the device, the device is a smartwatch containing at least two microphones.

In a possible embodiment of the device the device is a wearable comprising at least two microphones.

These and other aspects, embodiments, and advantages of illustrative embodiments will become apparent from the embodiments described herein when taken in conjunction with the accompanying drawings. It is to be understood, however, that the specification and drawings are constructed for the purpose of illustration only and are not designed as a definition of the limits of the disclosed invention. Reference should be made to the appended claims for the disclosed invention. Additional aspects and advantages of the invention will be set forth in the description that follows, and in part will be obvious from the description, or may be learned by practice of the invention. Moreover, the aspects and advantages of the invention may be realized and obtained by means of the instrumentalities and combinations particularly pointed out in the appended claims.

In the following detailed portion of the disclosure, the invention will be described in greater detail with reference to exemplary embodiments illustrated in the following drawings.

1 shows a schematic perspective view of an exemplary apparatus incorporating aspects of the disclosed embodiment; FIG. 4 illustrates an exemplary method incorporating aspects of the disclosed embodiment; 4 illustrates aspects of an exemplary method incorporating aspects of the disclosed embodiment; 4 illustrates aspects of an exemplary method incorporating aspects of the disclosed embodiment; FIG. 5 is a graph showing the results of the wind detection method of the disclosed embodiments; FIG. 4A-4D illustrate any exemplary method aspects incorporating aspects of the disclosed embodiments; 1 is a schematic block diagram of an exemplary apparatus that can be used to implement aspects of the disclosed embodiment; FIG.

Referring to FIG. 1, there is shown a schematic block diagram of an exemplary apparatus 100 incorporating aspects of the disclosed embodiments. An aspect of the disclosed embodiments is the wind in one or more microphones of a wearable-like device or device such as a headset, mobile communication device, smartwatch, etc., comprising at least two microphones or audio signal processing devices. The purpose is to detect noise. Although aspects of the disclosed embodiments are generally described herein in the context of headsets or mobile communication devices, aspects of the disclosed embodiments are not so limited. Aspects of the disclosed embodiments may also be applied to detection of handling noise, such as when the device's microphone is touched by the user. In alternative embodiments, wind detection may be applied to any speech or audio signal processing device where wind or handling noise can affect the quality of the microphone signal.

As shown in FIG. 1, exemplary apparatus 100 may include at least one processor 102 and at least two audio signal input devices 104a, 104b. In other embodiments, apparatus 100 may include any suitable number of audio signal input devices 104a-104b other than including two. Aspects of the disclosed embodiment are not limited by the number of audio signal input devices 104a-104n exceeding two. For the purposes of this description, audio signal input devices 104a-104n are referred to as microphones.

Referring to FIG. 1, P _m denotes the power of the microphone signal S _m and P _b denotes the power of the beamformed signal at two or more of the microphones 104a-104n. It is understood that in some cases, beamformed signals may be computed from some, but not all, of the microphones 104a-104n.

Generally, microphones 104a-104n are configured to generate respective microphone signals Sm1 _{- Smn} in response to received signals, such as audio or noise. In one embodiment, the processor 102 may be configured to determine and/or calculate the powers _{Pm1- Pmn of} one or more of the respective microphone signals Sm1 _-Smn . For the purposes of this description, the power _Pm of the microphone signal _Sm will be referred to as the microphone signal power _Pm , and the power _Pb of the beamforming signal will be referred to as the beamforming signal power _Pb .

Wind noise can have a detrimental effect on one or more of the microphone signals S _m1 to S _mn , such as when the device 100 is used outdoors. This wind noise interference can hinder or otherwise degrade the ability to communicate with mobile communication devices.

In one embodiment, the power _Pm of microphone signal Sm is calculated _, such as the power _Pm1 of microphone signal _Sm1 of microphone 104a. The power P _m1 of the microphone signal S _m1 of the microphone 104a is then compared with the beamforming power signal P _b calculated from the at least two microphone signals. Wind noise can also be detected in one or more of the microphone signals of the device 100 based on the comparison.

In one embodiment, high values of the ratio P _m /P _b may indicate ambient noise or directional noise (no wind noise). A value of 1 for the ratio P _m /P _b may indicate speech from the target direction (no wind noise). Small values of the ratio P _m /P _b may indicate wind noise and also handling noise.

Aspects of the disclosed embodiment detect wind noise by calculating the ratio between the microphone signal power P _m to the beamforming signal power P _b and comparing the result to a threshold. As shown in the example of FIG. 1, when more than one microphone signal S _m1 to S _mn is present, the determination and use of the corresponding microphone signal power P _m1 to P _mn enhances wind noise detection. do. The beamformed signal power _Pb keeps audio signals such as speech in the target direction unchanged and attenuates sounds coming from other directions.

In one embodiment, filter-and-sum beamforming may be utilized to generate the beamformed signal power _Pb . Here, the two or more microphone signals S _m1 through S _mn (generally referred to herein as microphone signals S _m1 through S _mn ) used to calculate the beamforming signal power P _b are , is filtered and then added. In the presence of wind noise, the amplitude and phase differences between the microphone signals S _m1 to S _mn may not be as expected. The phase and amplitude relationship between the microphone signals S _m1 to S _mn depends on the direction from which the sound reaches the microphones 104a to 104n, also referred to herein as the microphone array. In the presence of wind noise, the relationship between amplitude and phase can change rapidly and can be very random. In some cases, the microphone signals S _m1 to S _mn may add, which may adversely affect proper audio signal processing.

Referring to FIG. 2, according to aspects of the disclosed embodiment, wind noise can be detected from the ratio of microphone signal power to beam power (P _m /P _b ). In one embodiment, the power P _m of the microphone signal S _m is determined 202 . The power _Pb of the beamformed signal is also determined. A ratio of P _m /P _b is calculated 206 . From this ratio, it is determined 208 whether there is wind noise. For example, a result of the ratio P _m /P _b of 1 indicates sound coming from the target direction without wind noise. If the ratio is greater than 1, this indicates ambient noise without wind noise. If the ratio is less than 1, it indicates wind noise.

If it is determined 208 that there is no wind noise 209, beamformer processing may be used on the microphone signals S _m1 to S _mn . In some cases, beamforming may be performed on at least two, but not all, of the microphone signals S _m1 to S _mn . If the determination 208 indicates the presence of wind noise 211, beamforming may be switched off 212. FIG. In one embodiment, the best microphone or the microphone that is exposed to the least amount of wind noise may be selected. In other embodiments, the microphone closest to the user's mouth may be selected. The selected microphone may be used 214 for further audio signal processing.

In one embodiment, wind noise detection is more accurate when the beamforming signal power P _b is compared to two or more microphone signal powers P _m1 to P _mn . The more microphone signal powers P _m1 to P _mn are used in the comparison, the more accurate the detection of wind noise.

For example, each of the microphone signal powers _Pm1 through _Pmn may be separately compared to the beamforming signal power _Pb . The determined lowest power ratio may be compared to a threshold. In other embodiments, the average value of total power ratios may be taken or used as an input, such as an input to a machine learning model.

In alternative embodiments, the beamformed power signal _Pb may be compared with both the filtered and unfiltered signal, or with the filtered signal only. The filtered signal may be, for example, a beamformer filter in filter-and-sum beamforming. Wind detection according to aspects of the disclosed embodiment can be further enhanced if the microphone signal power P _m is averaged over a period of several seconds.

Referring to FIG. 3, the result of the ratio P _m /P _b is compared 302 with the threshold thr. In this example, if the value of P _m /P _b is greater than 304 thr, this indicates the absence 306 of wind noise in microphone signal S _m or the absence of interfering wind noise. there is If the value of P _m /P _b is not greater than thr 308, this indicates that there is wind noise 310 in the microphone signal S _m .

Referring to FIG. 4, in one embodiment, wind noise detection is performed by calculating the microphone signal power to beam power ratio (P _m /P _b ) in frequency bands and then choosing the minimum value over these bands. is strengthened. In the example of FIG. 4, _Pm is the microphone signal power, _Pb is the beamformer power, thr is the threshold, f is the frequency band, and F is the set of frequency bands.

As shown in the example of FIG. 4, the values of P _m (f), P _b (f) are calculated 402 for all f in the set F of frequency bands. The values of P _m (f)/P _b (f) are calculated 404 for all frequency bands f in the set F of frequency bands. A value of R _min =min(P _m (f)/P _b (f)) is calculated 406 for every frequency band f in the set F of frequency bands. It is determined 408 whether the value of R _min is greater than thr. If R _min >thr, this indicates 410 that there is no wind or any interfering wind. If R _min <thr, this indicates the presence of wind 412, or the presence of an interfering wind.

Referring to FIG. 5, in one embodiment, wind noise detection can also be performed using two thresholds. In this example, wind noise is detected when the microphone signal-to-beam power ratio P _m /P _b is less than the lower threshold thr2, and detection is made when the signal-to-beam power ratio P _m /P _b is higher. It is canceled only when it is greater than the threshold thr1. The setting of thresholds thr1 and thr2 may depend on the characteristics and aspects of device 100 . Some of these features and aspects may include, but are not limited to, microphone positions, beam designs, and other parameter values. In general, the thresholds thr1 and thr2 may be designed to vary approximately between the ranges of 0 and -15 dB.

The example of FIG. 5 shows the microphone signal power to beam power ratio calculated from data measured with two microphones 104a, 104b approximately 1.2 cm apart. The minimum value of the frequency band R _min is plotted as solid curve 502 . This is also averaged over time. The averaging is Xave(i+1)=Xframe+a*(Xave(i)-Xframe). where Xave is the average value, Xframe is the value calculated for one frame, and a is a constant that determines how slow the averaging is (eg, a=0.99). Two thresholds, thr1 and thr2, are plotted with dashed lines 504,506. There is wind until about 31 seconds later, and then there is no wind.

In one embodiment, the values calculated in the above examples, such as the microphone signal power to beam power ratio _Pm / _Pb or _Rmin , are used as features in a machine learning algorithm, either alone or in combination with other features. May be used in combination.

Beamforming usually increases microphone noise at low frequencies. Under silent or non-interfering wind conditions, the beam power _Pb can be higher than the microphone signal power _Pm . In this example, the ambient noise level N is evaluated 610 and wind noise detection is bypassed if the noise level N is less than some threshold (thrn).

FIG. 6 shows a wind noise detection algorithm using noise level estimation. As shown in the example of FIG. 6, the values of P _m (f), P _b (f) are calculated 602 for all f in the set F of frequency bands. The values of P _m (f), P _b (f) are calculated 604 for all frequency bands f in the set F of frequency bands. A value of R _min =min(P _m (f)/P _b (f)) is calculated 606 for every frequency band f in the set F of frequency bands. It is determined 608 whether the value of R _min is greater than thr. If R _min is greater than thr, this indicates 620 that there is no wind or any interfering wind.

If R _min is not greater than thr, then it is determined 612 whether the noise level N is greater than a threshold thrn. During wind noise, the ambient noise level N increases. Therefore, whenever there is wind, wind noise is detected. Any commonly used noise estimation 610 method may be used for this task.

If the value of the ambient noise level N is not greater than the threshold thrn, it indicates 620 that there is no ambient noise or wind noise. A value of ambient noise level N greater than the threshold thrn indicates the presence 622 of ambient or wind noise.

In one embodiment, when wind noise is detected, the effects of the wind noise can be reduced, for example, by switching off the beamforming of the device 100 . It may be determined which of the microphones 104a-104n are exposed to the least amount of wind noise. In one embodiment, determining which of the microphones 104a-104n is exposed to the least amount of noise is by comparing microphone signal power P _m levels in a particular frequency band, such as below 3500 Hz. including. The microphone 104a-104n with the lowest power level _Pm during wind noise has the least amount of wind. The microphone that is exposed to the least amount of wind noise may then be selected for further operation and audio signal processing. Beamforming is used for audio signal processing when it is determined that there is no wind noise. In alternate embodiments, any manner of suitable wind noise reduction may be implemented when wind noise is detected in accordance with aspects of the disclosed embodiments.

FIG. 7 shows a block diagram of an exemplary apparatus 1000 suitable for implementing aspects of the disclosed embodiments. Apparatus 1000 is suitable, for example, for use in a wireless communication network.

Device 1000 includes or is coupled to processor or computing hardware 1002 , memory 1004 , radio frequency (RF) unit 1006 and user interface (UI) 1008 . In some embodiments, device 1000 does not include UI 1008 . Apparatus 1000 may also include two or more sound processing devices 1014, also referred to as microphones. In one embodiment, the two or more sound processing devices 1014 comprise arrays of at least two microphones as described with respect to FIG. Microphone 1014 is shown coupled to UI 1008, but may be connected to or within the device in any suitable manner. In one embodiment, device 1000 includes device 100 referenced in FIG.

Processor 1002 may be a single processing device, or may be, for example, a digital signal processing (DSP) device, a microprocessor, a graphics processing unit (GPU), a specialized processing device, or a general purpose computer processing device (CPU). may include a plurality of processing devices, including special purpose devices. Processor 1002 may be implemented as processor 102 described with respect to FIG. 1 and may be configured to implement any one or more of the methods and processes described herein.

In the example of FIG. 7, a combination of various types of volatile and nonvolatile computer memory, such as read only memory (ROM), random access memory (RAM), magnetic or optical disks, or other types of computer memory. Processor 1002 is configured to be coupled to memory 1004, which may be . Memory 1004 is configured to store computer program instructions that processor 1002 may access and execute to cause processor 1002 to perform various desired computer-implemented processes or methods, such as the methods described herein. . Memory 1004 may be implemented as part of or in combination with device 100 described with respect to FIG.

The program instructions stored in memory 1004 may be organized as sets or groups of program instructions, which are referred to in the industry using various terms such as program, software component, software module, and unit. Each module contains a set of functionality designed, alone or in combination, to support a specific purpose of carrying out one or more aspects of the aspects of the disclosed embodiments described herein. good. Also included in memory 1004 are program data and data files that may be stored and processed by processor 1002 while executing a set of computer program instructions.

Further, apparatus 1000 may include or be coupled to an RF unit 1006, such as a transceiver. RF unit 1006 is coupled to processor 1002 and configured to transmit and receive RF signals based on digital data 1012 exchanged with processor 1002 . It may be configured to send and receive wireless signals with other nodes in the wireless network. To facilitate transmission and reception of RF signals, RF unit 1006 includes an antenna unit 1010, which may include multiple antenna elements in certain embodiments. Multiple antennas 1010 may be configured to support transmission and reception of MIMO signals, as may be used for beamforming.

UI 1008 may include one or more user interface elements such as a touch screen, keypad, buttons, voice command processor, and other elements adapted to exchange information with a user. UI 1008 may also include display units configured to display various information suitable for computing devices or mobile user equipment, such as organic light emitting diodes (OLED), liquid crystal displays (LCD), and LED or using any suitable display type, such as less complex elements such as indicator lamps.

Aspects of the disclosed embodiment provide new methods for detecting wind noise. The presence of wind noise is determined by calculating the ratio between the microphone signal power and the beamforming signal power and comparing the ratio to a threshold. Any mobile device used outside will suffer from wind noise, and in the worst case wind noise will completely destroy the microphone signal, making it impossible to make a phone call, for example. Aspects of the disclosed embodiment can reliably detect wind noise. Once wind noise is detected, wind noise interference can be reduced using any one of a number of known methods, such as switching off beamforming and using single microphone processing. Also, internal sensors of the headset during wind noise, such as VACC (voice accelerometer), VPU (voice pickup sensor) or internal microphone) may be utilized. It also avoids false positives that can degrade the quality of phone calls and other audio features. The methods of the disclosed embodiments are also computationally efficient and easy to combine with other processing in devices such as mobile communication devices.

Thus, while the basic novel features of the invention have been shown, described and pointed out as applied to the exemplary embodiments, variations in form and detail of the illustrated devices and methods, as well as their operation, have been made. It is understood that omissions, substitutions and modifications may be made by those skilled in the art without departing from the spirit and scope of the presently disclosed invention. Moreover, it is expressly intended that all combinations of those elements which perform substantially the same function in substantially the same manner to achieve the same results are within the scope of the invention. Furthermore, the structures and/or elements shown and/or described in connection with any disclosed form or embodiment of the invention may, as a general matter of design choice, be incorporated in any other disclosure or description, or may be incorporated into any proposed form or embodiment. It is the intention, therefore, to be limited only as indicated by the claims appended hereto.

In a possible embodiment of the apparatus, the processor calculates a ratio of microphone signal power to beamforming signal power over a plurality of frequency bands, and calculates the calculated ratio over a plurality of frequency bands for comparison with a predetermined threshold. It is further configured to select a minimum value of the power ratio . Wind noise detection is enhanced by calculating the signal-to-beam power ratio over frequency bands and selecting the minimum value over these bands.

In a possible embodiment of the method, the comparing step includes: calculating a ratio of microphone signal power to beamforming signal power over a plurality of frequency bands; selecting a value; and comparing the selected minimum value to a predetermined threshold to determine wind noise. Wind noise detection is enhanced by calculating the signal-to-beam power ratio over frequency bands and selecting the minimum value over these bands.

In a possible embodiment of the method the device is a smartwatch comprising at least two microphones.

In a possible embodiment of the method the device is a wearable comprising at least two microphones.

Aspects of the disclosed embodiment detect wind noise by calculating the ratio of the microphone signal power P _m to the beamforming signal power P _b and comparing the result to a threshold. As shown in the example of FIG. 1, when more than one microphone signal S _m1 to S _mn is present, the determination and use of the corresponding microphone signal power P _m1 to P _mn enhances wind noise detection. do. The beamformed signal power _Pb keeps audio signals such as speech in the target direction unchanged and attenuates sounds coming from other directions.

Aspects of the disclosed embodiment provide new methods for detecting wind noise. The presence of wind noise is determined by calculating the ratio between the microphone signal power and the beamforming signal power and comparing the ratio to a threshold. Any mobile device used outside will suffer from wind noise, and in the worst case wind noise will completely destroy the microphone signal, making it impossible to make a phone call, for example. Aspects of the disclosed embodiment can reliably detect wind noise. Once wind noise is detected, wind noise interference can be reduced using any one of a number of known methods, such as switching off beamforming and using single microphone processing. Also, internal sensors of the headset during wind noise such as VACC (voice accelerometer), VPU (voice pickup sensor) or internal microphone may be utilized. It also avoids false positives that can degrade the quality of phone calls and other audio features. The methods of the disclosed embodiments are also computationally efficient and easy to combine with other processing in devices such as mobile communication devices.

Claims (15)

An apparatus (100) comprising a processor (102), said processor (102) comprising:
determining the power (P _m1 ) of the microphone signal (S _m1 ) of at least one microphone (104a) of the device;
determining the power (P _b ) of the beamformed signals of at least two microphones (104a, 104n) of the device;
comparing the microphone signal power (P _m1 ) to the beamforming signal power (P _b );
A device configured to detect wind noise in a microphone signal of the device (100) based on the comparison. The processor (102)
comparing the microphone signal power (P _m1 ) to the power of the beamformed signal (P _b ) by calculating the ratio of the microphone signal power (P _m1 ) to the beamformed signal power (P _b );
The apparatus (100) of claim 1, further configured to detect the wind noise based on the calculated ratio. The apparatus (100) of claim 2, wherein the processor (102) is further configured to detect the wind noise when the calculated ratio is less than a predetermined threshold. The processor (102)
switching off beamforming of the at least two microphones (104a, 104b) when wind noise is detected;
4. Further configured to select, for further audio signal processing, the microphone of the at least two microphones (104a, 104b) for which the least amount of wind noise is detected. The apparatus (100) according to any one of Claims 1 to 3. The processor (102)
power (P m2 ) of at least a second microphone signal (S _m2 ) of at least one other microphone (104b _{) of said at least two microphones (104a, 104b) as said beamforming signal power (P b )} ; compare and
further configured to detect the wind noise based on comparing the microphone signal power (P _m1 ) and the microphone signal power (P _m2 ) with the beamforming signal power (P _b ); A device (100) according to any one of claims 1 to 4. 6. The method of any of claims 1-5, wherein the processor (102) is further configured to calculate a ratio of the microphone signal power ( _Pm1 ) to the beamforming signal power ( _Pb ) over a frequency band. A device (100) according to any one of the preceding claims. The processor (102) calculates a ratio of the microphone signal power ( _Pm1 ) to the beamformed signal power ( _Pb ) over a plurality of frequency bands and compares the ratio to a predetermined threshold value for a plurality of frequency bands. 7. The apparatus (100) of claim 6, further configured to select a minimum value ( _Rmin ) of the ratio of powers calculated over a period of time. The apparatus (100) according to any one of claims 1 to 7, wherein the apparatus (100) is a mobile communication device. A method (200) for detecting wind noise in a microphone signal of a device, comprising:
determining (202) the power ( _Pm1 ) of the microphone signal ( _Sm1 ) of at least one microphone of said device;
determining (204) the power (P _b ) of the beamformed signals of at least two microphones of the device;
comparing (206) the microphone signal power (P _m1 ) with the beamforming signal power (P _b );
and detecting (208) wind noise in the device's microphone signal based on the comparison. By calculating 302 the ratio of the microphone signal power (P _m1 ) to the beamformed signal power (P _b ), the power (P _m1 ) of the microphone signal (S _m1 ) is divided into the beamformed signal power ( P _b ), comparing (206);
10. The method (200) of claim 9, further comprising detecting (208) the wind noise in the microphone signal based on the calculated ratio. 11. The method (200) of any one of claims 10, further comprising detecting (208) the wind noise if the calculated ratio is less than a predetermined threshold (308). . switching off beamforming of the at least two microphones (212) when wind noise is detected (208);
selecting (214) the microphone of the at least two microphones for which the least amount of wind noise is detected for further audio signal processing. 2. The method (200) of claim 1. comparing (206) the power (P _m2 ) of at least one other microphone signal (S _m2 ) of said at least one microphone of said device with said beamforming signal power (P _b );
detecting (208) the wind noise based on comparing the microphone signal power ( _Pm1 ) and the microphone signal power ( _Pm2 ) with the beamforming signal power ( _Pb ); A method (200) according to any one of claims 9 to 12, comprising: 14. The step of comparing (206) further comprising calculating (404) the ratio of the microphone signal power ( _Pm1 ) to the beamforming signal power ( _Pb ) over a frequency band. The method (200) of any one of Claims 1 to 3. The step of comparing (206) comprises:
calculating (402) a ratio of the microphone signal power ( _Pm1 ) to the beamforming signal power ( _Pb ) over a plurality of frequency bands;
selecting (406) the minimum value of the calculated power across a plurality of frequency bands;
15. The method (200) of any one of claims 9 to 14, further comprising comparing (408) the selected minimum value to a predetermined threshold to determine (208) the wind noise. ).

Images