Speaker location method, device and equipment

Abstract

The present application discloses a speaker localization method and device. The method collects multiple sound signals through a microphone array; divides the sound source range into multiple sound source positions; obtains the sound signal transmission factor from the sound source position to each microphone position according to the microphone array morphology information and the sound source position; forms a sound signal transmission model according to the transmission factor, the sound source signal and the multiple sound signals; and selects at least one target sound source position from the multiple sound source positions according to the transmission model, and the target sound source position includes the speaker's direction information and distance information. This processing method makes it possible to jointly estimate the speaker's direction and distance in a multi-speaker scene, avoiding the optimization of the speaker's direction or distance alone, so that the speaker's direction and distance can be better taken into account, and at the same time, it is compatible with different microphone directivities and multi-speaker localization scenes.

Description

Technical Field

[0001] The present application relates to the field of speech processing technology, and in particular to a speaker positioning method and device, and a conference terminal. Background Art

[0002] Speaker localization is widely used in voice communication, such as for beamforming for high-quality sound pickup, and for distance estimation for video scaling. Sound source localization for multiple speakers in audio and video communication systems has always been one of the problems that the industry needs to solve urgently.

[0003] Important indicators in positioning information include the direction-of-arrival (DOA) and the distance (Distance) of the sound source. Generally speaking, since it is assumed that the speech signal collected by the microphone array comes from the far-field model, more attention is paid to the DOA estimation problem. At present, DOA estimation mainly adopts spatial spectrum estimation (such as multi-signal classification method MUSIC, steerable beam response SRP-PHAT), etc. Compared with DOA estimation, sound source distance estimation has not received much attention in the industry.

[0004] However, the inventors found that the existing solutions have at least the following problems: the estimation method based on the near-end signal transmission model to improve the DOA estimation solution cannot take into account the DOA angle estimation. In particular, the current DOA estimation is mostly based on omnidirectional microphone arrays, and the application of directional microphones is not considered. Summary of the invention

[0005] The present application provides a speaker positioning method to solve the problems in the prior art that the speaker direction and distance cannot be well considered and different types of microphones cannot be compatible. The present application also provides a speaker positioning device and a conference terminal.

[0006] The present application provides a speaker localization method, comprising:

[0007] Collect multiple sound signals through microphone array;

[0008] Divide the sound source range into multiple sound source positions;

[0009] According to the microphone array morphology information and the sound source position, obtaining a sound signal transmission factor from the sound source position to each microphone position;

[0010] forming a sound signal transmission model according to the transmission factor, the sound source signal and the multi-channel sound signals;

[0011] According to the transmission model, at least one target sound source position is selected from the multiple sound source positions, and the target sound source position includes direction information and distance information of a speaker.

[0012] Optionally, the microphone array is an omnidirectional microphone array; and obtaining the sound signal transmission factor from the sound source position to each microphone position according to the microphone array morphology information and the sound source position includes:

[0013] Acquire angle information between the sound source position and the microphone position according to the microphone array morphology information and the sound source position;

[0014] According to the angle information, obtaining distance information between the sound source position and the microphone position;

[0015] According to the distance information between the sound source position and the microphone position, acquiring the attenuation information and the phase shift information of the sound source signal propagating to the microphone position;

[0016] The sound signal transmission factor is formed according to the attenuation information and the phase shift information.

[0017] Optionally, the microphone array is a directional microphone array; and obtaining the sound signal transmission factor from the sound source position to each microphone position according to the microphone array morphology information and the sound source position includes:

[0018] Acquire angle information between the sound source position and the microphone position according to the microphone array morphology information and the sound source position;

[0019] According to the angle information, obtaining distance information between the sound source position and the microphone position;

[0020] According to the distance information between the sound source position and the microphone position, acquiring the attenuation information and the phase shift information of the sound source signal propagating to the microphone position;

[0021] Acquire amplitude response information between the sound source position and the microphone position according to the angle information;

[0022] The sound signal transmission factor is obtained according to the attenuation information, the phase shift information and the amplitude response information.

[0023] Optionally, also include:

[0024] Get the noise signal of the microphone;

[0025] The forming of a sound signal transmission model according to the transmission factor, the sound source signal and the multi-path sound signal comprises:

[0026] Taking the product of the transmission factor and the sound source signal as the attenuated sound source signal;

[0027] The multi-channel sound signal is a sum of a plurality of attenuated sound source signals and the noise signal as the sound signal transmission model.

[0028] Optionally, selecting at least one target sound source position from the multiple sound source positions according to the transmission model includes:

[0029] The transmission model is used as prior knowledge for speaker localization, and the sound source signal is used as a latent variable. At least one target sound source position is selected from the multiple sound source positions through a sparse Bayesian learning method.

[0030] Optionally, selecting at least one target sound source position from the multiple sound source positions by using a sparse Bayesian learning method includes:

[0031] The maximum expectation algorithm is adopted to select at least one target sound source position information from the multiple sound source positions through a sparse Bayesian learning method.

[0032] Optionally, selecting at least one target sound source position from the multiple sound source positions by using a sparse Bayesian learning method includes:

[0033] Selecting at least one first target sound source position in the three-dimensional spatial spectrum from the multiple sound source positions by a sparse Bayesian learning method;

[0034] The target sound source position in the three-dimensional spatial spectrum is mapped to the two-dimensional spatial spectrum to obtain at least one second target sound source position in the two-dimensional spatial spectrum.

[0035] Optionally, the sound signal includes sound signals of multiple speakers.

[0036] The present application also provides a speaker location device, comprising:

[0037] A signal collection unit, used for collecting multi-channel sound signals through a microphone array;

[0038] A sound source range division unit, used for dividing the sound source range into a plurality of sound source positions;

[0039] A transmission factor acquisition unit, used for acquiring a sound signal transmission factor from the sound source position to each microphone position according to the microphone array morphology information and the sound source position;

[0040] A transmission model acquisition unit, used to form a sound signal transmission model according to the transmission factor, the sound source signal and the multi-channel sound signal;

[0041] The joint estimation unit is used to select at least one target sound source position from the multiple sound source positions according to the transmission model, and the target sound source position includes the direction information and distance information of the speaker. The present application also provides a computer-readable storage medium, which stores instructions, and when it is run on a computer, the computer executes the above-mentioned various methods.

[0042] The present application also provides a computer program product comprising instructions, which, when executed on a computer, enables the computer to execute the above-mentioned various methods.

[0043] Compared with the prior art, this application has the following advantages:

[0044] The speaker localization method provided in the embodiment of the present application collects multiple sound signals through a microphone array; divides the sound source range into multiple sound source positions; obtains the sound signal transmission factor from the sound source position to each microphone position according to the microphone array morphology information and the sound source position; forms a sound signal transmission model according to the transmission factor, the sound source signal and the multiple sound signals; and selects at least one target sound source position from the multiple sound source positions according to the transmission model, and the target sound source position includes the speaker's direction information and distance information. This processing method makes it possible to jointly estimate the speaker's direction and distance in a multi-speaker scene, avoiding the optimization of the speaker's direction or distance alone, so that the speaker's direction and distance can be better taken into account, and at the same time, it is compatible with different microphone directivities and multi-speaker localization scenes. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 A flowchart of an embodiment of a speaker localization method provided by the present application;

[0046] Figure 2 A schematic diagram of sound signal transmission of an embodiment of a speaker localization method provided by the present application;

[0047] Figure 3 Schematic diagram of sparse Bayesian learning of an embodiment of the speaker localization method provided by the present application;

[0048] Figure 4 A schematic diagram of spatial spectrum mapping of an embodiment of a speaker localization method provided in the present application. DETAILED DESCRIPTION

[0049] Many specific details are described in the following description to facilitate a full understanding of the present application. However, the present application can be implemented in many other ways than those described herein, and those skilled in the art can make similar generalizations without violating the connotation of the present application, so the present application is not limited by the specific implementation disclosed below.

[0050] In the present application, a speaker location method and device, and a conference terminal are provided. In the following embodiments, various solutions are described in detail one by one.

[0051] First embodiment

[0052] Please see Figure 1 , which is a flow chart of an embodiment of the speaker localization method of the present application. In this embodiment, the method may include the following steps:

[0053] Step S101: Collect multiple sound signals through a microphone array, wherein the multiple sound signals include two or more sound signals.

[0054] The method can be applied to conference terminal equipment in an audio and video conference system, and is suitable for locating or distinguishing two or more people in different locations. The audio and video conference system transmits the voice, image and file data of two or more individuals or groups in different locations to each other through conference terminals and transmission lines, etc., to achieve instant and interactive communication, so as to achieve efficient and instant conference communication.

[0055] In a single-speaker scenario, sound source localization only needs to determine the speaker's direction (DOA, direction of arrival / angle). In a multi-speaker scenario, there may be two sound sources at the same angle, and at this time, distance information (front and back) is required. Based on distance information, more accurate sound source localization can be achieved in multi-speaker scenarios. The method provided in the embodiment of the present application can jointly estimate the direction of arrival (DOA) and distance of multiple near-field sound sources.

[0056] The conference terminal includes a microphone array. The microphone array can collect sound signals of the local conference site, and the microphone array can collect multi-channel sound signals. From the perspective of microphone type, the microphone array can be an array composed of omnidirectional microphones, or an array composed of directional microphones. From the perspective of array shape, the microphone array can be a linear array or a circular array, or an array of irregular shape.

[0057] Microphone directivity is used to indicate the sensitivity of a microphone to sounds coming from all directions in space, and is an important attribute of a microphone. For example, for a general omnidirectional microphone, it has the same sensitivity to sounds coming from different angles, while for other types of microphones such as cardioid, hypercardioid, and bi-directional, the sensitivity to sounds coming from different angles varies.

[0058] Step S103: Divide the sound source range into multiple sound source positions.

[0059] The method provided in the embodiment of the present application jointly estimates the direction information and distance information of multiple speakers, and needs to discretize the sound source range into multiple potential sound source positions. Then, at least one target sound source position can be determined from the multiple potential sound source positions, that is, the direction information and distance information of multiple speakers.

[0060] In this embodiment, the near-field sound source signal carries distance information, and the near-field sound source range is divided into multiple potential sound source positions. For example, the range space of about 5 to 6 meters away from the microphone array is divided into hundreds of potential sound source positions, which can be recorded as , where K represents the number of potential sound source positions divided, such as K=640 means that the sound source range is divided into 640 sound source positions, and the subsequent steps will determine at least one target sound source position from these 640 potential sound source positions.

[0061] In specific implementation, the sound source range can be divided into multiple sound source positions in any manner, such as divided into multiple rectangles with smaller areas, and a position in the rectangle (such as the midpoint) is selected as the sound source position represented by the rectangle, or divided into other densely tiled shapes, such as triangles or regular hexagons.

[0062] Step S103: acquiring a sound signal transmission factor from the sound source position to each microphone position according to the microphone array morphology information and the sound source position.

[0063] The transmission factor is related to the transmission path between the sound source position and the microphone position. The transmission path includes the distance and angle between the sound source position and the microphone position. The transmission path from the same sound source position to different microphones in the microphone array is related to the microphone array morphology. Taking a linear directional microphone array as an example, the array morphology information may include the distance between the microphones, and the microphone pointing direction may be perpendicular to the array pointing to one side. Taking a circular directional microphone array as an example, the microphone array morphology data may include the circular array radius r, the angle between the microphone and the horizontal direction of the microphone array In addition, the parameter of the transmission factor can also be a sound source identifier, through which multiple sound source scenarios can be covered.

[0064] In one example, the microphone array is an omnidirectional microphone array, and the transmission factor represents attenuation and phase shift during signal propagation. Step S105 may include the following sub-steps:

[0065] Step S201: acquiring angle information between the sound source position and the microphone position according to the microphone array morphology information and the sound source position.

[0066] In this embodiment, the microphone array may be composed of a plurality of (eg, 2M+1) microphones, and the distance between the microphones is d. Figure 2 As shown, the microphone in the middle position is used as the reference microphone, and the distance between the i-th sound source position and the reference microphone is r i , angle θ i , the distance between the mth microphone and the reference microphone is md, and the angle information between the ith sound source position and the mth microphone can be expressed by the following formula:

[0067]

[0068] Step S203: Acquire distance information between the sound source position and the microphone position according to the angle information.

[0069] In this embodiment, the distance information between the i-th sound source position and the m-th microphone can be expressed by the following formula:

[0070]

[0071] Step S205: acquiring attenuation information and phase shift information of the sound source signal propagating to the microphone position according to the distance information between the sound source position and the microphone position.

[0072] Step S207: forming the sound signal transmission factor according to the attenuation information and the phase shift information.

[0073] In this embodiment, the sound signal transmission factor from the i-th sound source position to the m-th microphone can be expressed by the following formula:

[0074]

[0075] in Represents attenuation information, Indicates phase shift information.

[0076] In one example, the microphone array is a directional microphone array, and the transmission factor represents not only the attenuation and phase shift during signal propagation, but also the amplitude response caused by the microphone directivity. Step S105 may include the following sub-steps:

[0077] Step S301: acquiring angle information between the sound source position and the microphone position according to the microphone array morphology information and the sound source position.

[0078] Step S303: Acquire distance information between the sound source position and the microphone position according to the angle information.

[0079] Step S305: acquiring attenuation information and phase shift information of the sound source signal propagating to the microphone position according to the distance information between the sound source position and the microphone position.

[0080] Step S307: Acquire amplitude response information between the sound source position and the microphone position according to the angle information.

[0081] In this embodiment, the amplitude response information between the i-th sound source position and the m-th microphone can be expressed by the following formula:

[0082]

[0083] Where α is the microphone shape parameter, α=1 represents an omnidirectional microphone, and 0<α<1 represents a directional microphone.

[0084] Step S309: Acquire the sound signal transmission factor according to the attenuation information, the phase shift information and the amplitude response information.

[0085] In this embodiment, the transmission factor includes the amplitude response caused by the directivity of the microphone and the phase and attenuation during the transmission process. The sound signal transmission factor from the i-th sound source position to the m-th microphone can be expressed by the following formula:

[0086] h m,i =g m,i a m,i

[0087] From the above formula, it can be seen that the transmission factor can cover the wave arrival direction of the sound source (θ i ), distance (r i ), number of sound sources, microphone directivity related parameters, etc.

[0088] In this embodiment, the near field range is discretized into K potential source positions. The expanded transmission factor matrix and source signal matrix can be expressed as:

[0089]

[0090]

[0091] Where H is a transfer factor matrix with 2M+1 (number of microphones) rows and K (number of divided sound source positions) columns, is an I-row sparse matrix, for each The non-zero rows are

[0092] Step S107: forming a sound signal transmission model according to the transmission factor, the sound source signal and the multi-channel sound signals.

[0093] The sound signal transmission model refers to the relationship between the sound signal collected by the microphone and multiple sound source signals, and may be related to the transmission factors between the multiple sound sources and each microphone.

[0094] In one example, step S107 can be implemented in the following manner: the product of the transmission factor and the sound source signal is used as the attenuated sound source signal; the multi-channel sound signal is the sum of multiple attenuated sound source signals as the sound signal transmission model. The following formula can be used to model the sound signal transmission (the sound signal transmission model):

[0095]

[0096] In another example, step S107 can be implemented in the following manner: the product of the transmission factor and the sound source signal is used as the attenuated sound source signal; the multi-channel sound signal is the sum of multiple attenuated sound source signals and the noise signal as the sound signal transmission model. The following formula can be used to model the sound signal transmission (the sound signal transmission model):

[0097]

[0098] Without considering the sparse sound source signal matrix In the case of zero row, the modeling of sound signal transmission (the sound signal transmission model) can adopt the following formula:

[0099]

[0100] In this formula, n is the time axis, ranging from 1 to the Nth moment; I is the number of sound sources. m represents the sound signal collected by the m-th microphone (ie, the m-th sound signal collected in step S01), which includes sounds from multiple sound sources; s i represents the signal of the i-th sound source; w m represents the noise of the mth microphone; h m,i represents the transmission factor between the i-th sound source and the m-th microphone; h m,i s i (n) represents the attenuated sound source signal of the i-th sound source collected by the m-th microphone.

[0101] Step S109: selecting at least one target sound source position from the multiple sound source positions according to the transmission model, wherein the target sound source position includes direction information and distance information of the speaker.

[0102] This step can be implemented by machine learning, which includes but is not limited to sparse Bayesian learning and neural network learning.

[0103] In a specific implementation, step S109 can be implemented in the following manner: using the transmission model as prior knowledge for speaker localization, using the sound source signal as a latent variable, and selecting at least one target sound source position from the multiple sound source positions through a sparse Bayesian learning method.

[0104] The method provided in the embodiment of the present application converts the parameters in the transmission factor: the direction of arrival and the distance (θ i , r i ) is used as the objective function and a sparse Bayesian learning method is used to estimate the multi-channel sound signals collected by the microphone.

[0105] like Figure 3 As shown, this embodiment adopts a sparse Bayesian learning method to estimate the direction of arrival and distance (θ i , r i ). The candidate sound source signal S is a latent variable, which a priori obeys a zero-mean complex Gaussian distribution. The distribution is the sound source location-related accuracy parameter γ = [γ1, γ2, …, γK]. The application of sparse Bayesian learning in multi-speaker localization scenarios can be expressed by the following probability formula:

[0106]

[0107] The prior of additive noise W obeys the zero-mean complex Gaussian distribution, and the precision parameter δ can be expressed by the following probability formula:

[0108]

[0109] The marginalized probability density function (PDF) can be expressed as:

[0110]

[0111] In the above probability formula, γ is the estimated parameter, including the direction of arrival and distance (θ i , r i ), δ represents the noise signal.

[0112] In one example, the obtaining of at least one target sound source position by sparse Bayesian learning may be implemented in the following manner: using a maximum expectation algorithm to obtain at least one target sound source position information by sparse Bayesian learning.

[0113] In order to estimate γ and δ, the expectation maximization algorithm (EM) can be used. The EM algorithm is an algorithm for finding the maximum likelihood estimate or maximum a posteriori estimate of parameters in a probability model that depends on unobservable hidden variables. The following formula can be used:

[0114]

[0115]

[0116]

[0117]

[0118] The above formula is used to estimate the parameters, which leads to faster convergence of large-scale and highly overcomplete problems.

[0119] In one example, step S109 can be implemented in the following manner: selecting at least one first target sound source position in the three-dimensional spatial spectrum from the multiple sound source positions through a sparse Bayesian learning method; mapping the target sound source position in the three-dimensional spatial spectrum to a two-dimensional spatial spectrum to obtain at least one second target sound source position in the two-dimensional spatial spectrum.

[0120] In a specific implementation, the expected value of the log-likelihood of the E-step and the parameter (γ, δ) estimation of the M-step in the above EM algorithm can be obtained by sparse Bayesian learning according to the multi-channel sound signals, the determined microphone selection and the array topology (H), and the parameter γ representing the direction of arrival and the distance can be obtained through multiple iterations (controlled by a set threshold). The speaker localization information of the estimated parameters is mapped to a two-dimensional spatial spectrum to obtain the direction information and distance information of the speaker.

[0121] Specifically, if Figure 4 As shown in the figure, after the speaker location information (parameter γ) is determined by the EM algorithm, the parameter can be mapped to the 2D spatial spectrum to determine the direction of arrival and distance (θ i , r i ).

[0122] As can be seen from the above embodiments, the speaker localization method provided in the embodiments of the present application collects multiple sound signals through a microphone array; divides the sound source range into multiple sound source positions; obtains the sound signal transmission factor from the sound source position to each microphone position according to the microphone array morphology information and the sound source position; forms a sound signal transmission model according to the transmission factor, the sound source signal and the multiple sound signals; and selects at least one target sound source position from the multiple sound source positions according to the transmission model, and the target sound source position includes the speaker's direction information and distance information. This processing method makes it possible to jointly estimate the speaker's direction and distance in a multi-speaker scene, avoiding the optimization of the speaker's direction or distance alone, so that the speaker's direction and distance can be better taken into account, and at the same time, it is compatible with different microphone directivities and multi-speaker localization scenes.

[0123] Second embodiment

[0124] In the above-mentioned embodiment, a speaker location method is provided. Correspondingly, the present application also provides a speaker location device. The device corresponds to the embodiment of the above-mentioned method. Since the device embodiment is basically similar to the method embodiment, the description is relatively simple. For the relevant parts, please refer to the partial description of the method embodiment. The device embodiment described below is only illustrative.

[0125] The present application further provides a speaker localization device, comprising: a signal collection unit, a sound source range division unit, a transmission factor acquisition unit, a transmission model acquisition unit, and a joint estimation unit.

[0126] A signal acquisition unit is used to collect multiple sound signals through a microphone array; a sound source range division unit is used to divide the sound source range into multiple sound source positions; a transmission factor acquisition unit is used to obtain the sound signal transmission factor from the sound source position to each microphone position according to the microphone array morphology information and the sound source position; a transmission model acquisition unit is used to form a sound signal transmission model according to the transmission factor, the sound source signal and the multiple sound signals; a joint estimation unit is used to select at least one target sound source position from the multiple sound source positions according to the transmission model, and the target sound source position includes the direction information and distance information of the speaker.

[0127] In a specific embodiment, the microphone array is an omnidirectional microphone array; the transmission factor acquisition unit is specifically used to obtain the angle information between the sound source position and the microphone position based on the microphone array morphology information and the sound source position; obtain the distance information between the sound source position and the microphone position based on the angle information; obtain the attenuation information and phase shift information of the sound source signal propagating to the microphone position based on the distance information between the sound source position and the microphone position; and form the sound signal transmission factor based on the attenuation information and the phase shift information.

[0128] In a specific embodiment, the microphone array is a directional microphone array; the transmission factor acquisition unit is specifically used to obtain the angle information between the sound source position and the microphone position based on the microphone array morphology information and the sound source position; obtain the distance information between the sound source position and the microphone position based on the angle information; obtain the attenuation information and phase shift information of the sound source signal propagating to the microphone position based on the distance information between the sound source position and the microphone position; obtain the amplitude response information between the sound source position and the microphone position based on the angle information; obtain the sound signal transmission factor based on the attenuation information, phase shift information and amplitude response information.

[0129] In a specific embodiment, the device may also include: a noise signal acquisition unit, used to acquire the noise signal of the microphone; the transmission model acquisition unit, specifically used to take the product of the transmission factor and the sound source signal as the attenuated sound source signal; and take the multi-channel sound signal as the sum of multiple attenuated sound source signals and the noise signal as the sound signal transmission model.

[0130] In a specific embodiment, the joint estimation unit is specifically used to use the transmission model as prior knowledge for speaker localization, use the sound source signal as a latent variable, and select at least one target sound source position from the multiple sound source positions through a sparse Bayesian learning method.

[0131] In a specific embodiment, the joint estimation unit is specifically configured to select at least one target sound source position information from the multiple sound source positions by adopting a maximum expectation algorithm through a sparse Bayesian learning method.

[0132] In a specific embodiment, the joint estimation unit is specifically used to select at least one first target sound source position in the three-dimensional spatial spectrum from the multiple sound source positions through a sparse Bayesian learning method; and map the target sound source position in the three-dimensional spatial spectrum to a two-dimensional spatial spectrum to obtain at least one second target sound source position in the two-dimensional spatial spectrum.

[0133] The voice signals include voice signals of multiple speakers.

[0134] Third embodiment

[0135] In the above-mentioned embodiment, a speaker location method is provided, and correspondingly, the present application also provides an electronic device. The device corresponds to the embodiment of the above-mentioned method. Since the device embodiment is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment. The device embodiment described below is only illustrative.

[0136] The present application further provides an electronic device, comprising: a microphone array; a processor; and a memory, wherein the memory is used to store a program for implementing any of the above methods, and the terminal is powered on and runs the program of the method through the processor.

[0137] The electronic device may be an audio or video conferencing terminal or a sound pickup device.

[0138] Although the present application is disclosed as above in the form of a preferred embodiment, it is not intended to limit the present application. Any technical personnel in this field may make possible changes and modifications without departing from the spirit and scope of the present application. Therefore, the scope of protection of the present application shall be based on the scope defined by the claims of the present application.

[0139] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0140] The memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0141] 1. Computer-readable media includes permanent and non-permanent, removable and non-removable media that can be used to store information by any method or technology. Information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include non-transitory media such as modulated data signals and carrier waves.

[0142] 2. Those skilled in the art should understand that the embodiments of the present application can be provided as methods, systems or computer program products. Therefore, the present application can take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

Claims

1. A speaker localization method, characterized in that: include: Collect multi-channel sound signals through microphone array; Divide the sound source range into multiple sound source positions; According to the microphone array morphology information and the sound source position, obtaining a sound signal transmission factor from the sound source position to each microphone position; forming a sound signal transmission model according to the transmission factor, the sound source signal and the multi-channel sound signals; The transmission model is used as prior knowledge for speaker localization, and the sound source signal is used as a latent variable. At least one target sound source position is selected from the multiple sound source positions through a sparse Bayesian learning method, and the target sound source position includes the direction information and distance information of the speaker.

2. The method according to claim 1, characterized in that The microphone array is an omnidirectional microphone array; The step of obtaining the sound signal transmission factor from the sound source position to each microphone position according to the microphone array morphology information and the sound source position includes: Acquire angle information between the sound source position and the microphone position according to the microphone array morphology information and the sound source position; According to the angle information, obtaining distance information between the sound source position and the microphone position; According to the distance information between the sound source position and the microphone position, acquiring the attenuation information and the phase shift information of the sound source signal propagating to the microphone position; The sound signal transmission factor is formed according to the attenuation information and the phase shift information.

3. The method according to claim 1, characterized in that The microphone array is a directional microphone array; The step of obtaining the sound signal transmission factor from the sound source position to each microphone position according to the microphone array morphology information and the sound source position includes: Acquire angle information between the sound source position and the microphone position according to the microphone array morphology information and the sound source position; According to the angle information, obtaining distance information between the sound source position and the microphone position; According to the distance information between the sound source position and the microphone position, acquiring the attenuation information and the phase shift information of the sound source signal propagating to the microphone position; Acquire amplitude response information between the sound source position and the microphone position according to the angle information; The sound signal transmission factor is obtained according to the attenuation information, the phase shift information and the amplitude response information.

4. The method according to claim 1, characterized in that Also includes: Get the noise signal of the microphone; The forming of a sound signal transmission model according to the transmission factor, the sound source signal and the multi-path sound signal comprises: Taking the product of the transmission factor and the sound source signal as the attenuated sound source signal; The multi-channel sound signal is a sum of a plurality of attenuated sound source signals and the noise signal as the sound signal transmission model.

5. The method according to claim 1, characterized in that The selecting at least one target sound source position from the multiple sound source positions by means of sparse Bayesian learning includes: The maximum expectation algorithm is adopted to select at least one target sound source position information from the multiple sound source positions through a sparse Bayesian learning method.

6. The method according to claim 1, characterized in that The obtaining of at least one target sound source position by sparse Bayesian learning includes: Obtaining at least one first target sound source position in the three-dimensional space spectrum by a sparse Bayesian learning method; The target sound source position in the three-dimensional spatial spectrum is mapped to the two-dimensional spatial spectrum to obtain at least one second target sound source position in the two-dimensional spatial spectrum.

7. The method according to claim 1, characterized in that The voice signals include voice signals of multiple speakers.

8. A speaker location device, characterized in that: include: A signal collection unit, used for collecting multi-channel sound signals through a microphone array; A sound source range division unit, used for dividing the sound source range into a plurality of sound source positions; A transmission factor acquisition unit, used for acquiring a sound signal transmission factor from the sound source position to each microphone position according to the microphone array morphology information and the sound source position; A transmission model acquisition unit, used to form a sound signal transmission model according to the transmission factor, the sound source signal and the multi-channel sound signal; The joint estimation unit is used to use the transmission model as prior knowledge for speaker localization and the sound source signal as a latent variable, and select at least one target sound source position from the multiple sound source positions through a sparse Bayesian learning method, wherein the target sound source position includes the direction information and distance information of the speaker.

9. A conference device, characterized in that: include: speaker; microphone; processor; as well as A memory is used to store a program for implementing the method according to any one of claims 1 to 7, and the terminal is powered on and runs the program of the method through the processor.

Images