A howling detection method, voice call method and related device

Abstract

The present application discloses a howling detection method, a voice call method and related devices, which can be applied to the voice intelligent processing process in cloud communication scenarios. By acquiring the audio input signal in the communication process, and extracting the periodic characteristics and spectral characteristics of the audio input signal, wherein the periodic characteristics include the autocorrelation peak intensity and the time difference position; then fusing the periodic characteristics and the spectral characteristics to obtain the fused characteristics; and then detecting the fused characteristics based on at least one detection condition to obtain the howling detection result. Thus, the howling detection process based on the fused characteristics is realized. Since the periodic characteristics can be used to trace the time when the early howling occurs, and the spectral characteristics after the howling is formed are combined to make the final judgment, the probability of false detection is reduced and the accuracy of howling detection is improved.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a howling detection method, a voice call method and related devices. Background Art

[0002] In audio communication, if two communication devices are close to each other, or if two devices in different places have echo leakage at the same time in long-distance communication, howling is likely to occur.

[0003] Generally, the howling detection process can start from the phenomenon of howling, that is, analyzing the spectrum characteristics of the howling sound. When the signal energy is concentrated in a narrow band, howling may exist, and howling detection is performed based on the spectrum characteristics of these narrow bands.

[0004] However, the process based on spectrum features has limitations. For example, it is easy to misdetect other naturally existing single-frequency narrowband signals such as telephone ringing and alarms. In addition, the detection process can only detect severe howling after it occurs, which affects the accuracy of howling detection. Summary of the invention

[0005] In view of this, the present application provides a howling detection method, which can effectively avoid howling misdetection caused by external factors and improve the accuracy of the howling detection process.

[0006] A first aspect of the present application provides a howling detection method, which can be applied to a system or program including a howling detection function in a terminal device, and specifically includes: obtaining an audio input signal;

[0007] Extracting periodic features of the audio input signal, wherein the periodic features include autocorrelation peak intensity and time difference position;

[0008] Extracting a frequency spectrum feature of the audio input signal, where the frequency spectrum feature is obtained by fusing frequency spectrum parameters of at least one dimension, and the frequency spectrum parameters are related to the frequency spectrum distribution of the howling point;

[0009] fusing the periodic feature and the spectrum feature to obtain a fused feature;

[0010] The fusion feature is detected based on at least one detection condition to obtain a howling detection result, wherein the detection condition is set based on the occurrence probability or intensity of the howling point.

[0011] Optionally, in some possible implementations of the present application, extracting the periodic feature of the audio input signal includes:

[0012] Determining an autocorrelation coefficient according to the repeatability of the audio input signal within a detection time period;

[0013] The autocorrelation peak intensity and the time difference position are determined according to the autocorrelation coefficient.

[0014] Optionally, in some possible implementations of the present application, determining the autocorrelation coefficient according to the repeatability of the audio input signal within the detection duration includes:

[0015] Obtaining a network transmission delay fluctuation range of the audio input signal within a collection time period;

[0016] Determining the detection duration according to the fluctuation range of the network transmission delay;

[0017] An autocorrelation coefficient is determined according to the repeatability of the audio input signal within the detection time period.

[0018] Optionally, in some possible implementations of the present application, determining the autocorrelation coefficient according to the repeatability of the audio input signal within the detection duration includes:

[0019] Processing the audio input signal based on a preprocessing rule, the preprocessing rule comprising determining noise characteristics in at least one dimension;

[0020] The autocorrelation coefficient is determined according to the repeatability of the processed audio input signal within the detection time period.

[0021] Optionally, in some possible implementations of the present application, extracting the frequency spectrum feature of the audio input signal includes:

[0022] Performing time-frequency analysis on the audio input signal to obtain a time-frequency signal;

[0023] Screening is performed according to the peak characteristics of the time-frequency signal to obtain a spectrum peak point;

[0024] Counting the occurrence probability of the peak point of the spectrum to obtain the peak point frequency;

[0025] The frequency spectrum feature is extracted according to the peak point frequency.

[0026] Optionally, in some possible implementations of the present application, extracting the spectrum feature according to the peak point frequency includes:

[0027] Acquire, according to the peak point frequency, an energy ratio of the frequency point energy of each frequency point in the time-frequency signal to the total energy of the time-frequency signal;

[0028] The energy stationarity of the time-frequency signal is determined according to the energy ratio to obtain the frequency spectrum feature.

[0029] Optionally, in some possible implementations of the present application, extracting the spectrum feature according to the peak point frequency includes:

[0030] Obtaining a harmonic signal corresponding to the time-frequency signal;

[0031] Acquire the frequency energy of each frequency point in the time-frequency signal according to the peak point frequency;

[0032] The frequency spectrum feature is determined according to the ratio of the frequency point energy to the energy corresponding to the harmonic signal.

[0033] Optionally, in some possible implementations of the present application, extracting the spectrum feature according to the peak point frequency includes:

[0034] Determine the frequency point corresponding to the peak point frequency;

[0035] The frequency stability is determined according to the fluctuation of adjacent frequency points among the frequency points corresponding to the peak point frequency, so as to obtain the frequency spectrum characteristics.

[0036] Optionally, in some possible implementations of the present application, the screening according to the peak characteristics of the time-frequency signal to obtain the spectrum peak point includes:

[0037] Determining candidate peak points according to the peak characteristics of the time-frequency signal;

[0038] The candidate peak points satisfying the preset conditions are selected to obtain the spectrum peak points.

[0039] Optionally, in some possible implementations of the present application, the fusing the periodic feature and the spectrum feature to obtain a fused feature includes:

[0040] Determining a weighting parameter according to the periodic feature, the weighting parameter being used to indicate the number of samples within a period in the periodic feature;

[0041] assigning weighting factors to the periodicity characteristic and the spectrum characteristic according to the weighting parameter;

[0042] The periodic feature after the weighting factor is assigned and the spectrum feature after the weighting factor is assigned are fused to obtain the fused feature.

[0043] Optionally, in some possible implementations of the present application, allocating a weighting factor to the periodic feature and the spectrum feature according to the weighting parameter includes:

[0044] Determine the network delay characteristics of the periodic characteristics and the spectrum characteristics according to the weighting parameters;

[0045] A weighting factor is assigned to the periodicity characteristic and the spectrum characteristic according to the network delay characteristic.

[0046] A second aspect of the present application provides a howling detection device, comprising: an acquisition unit, configured to acquire an audio input signal;

[0047] An extraction unit, used to extract the periodic features of the audio input signal, wherein the periodic features include autocorrelation peak intensity and time difference position;

[0048] The extraction unit is further used to extract a spectrum feature of the audio input signal, where the spectrum feature is obtained by fusing spectrum parameters of at least one dimension, and the spectrum parameter is related to the spectrum distribution of the howling point;

[0049] A fusion unit, used for fusing the periodic feature and the spectrum feature to obtain a fusion feature;

[0050] A detection unit is used to detect the fusion feature based on at least one detection condition to obtain a howling detection result, wherein the detection condition is set based on the occurrence probability or intensity of the howling point.

[0051] Optionally, in some possible implementations of the present application, the extraction unit is specifically configured to determine an autocorrelation coefficient according to the repeatability of the audio input signal within a detection duration;

[0052] The extraction unit is specifically configured to determine the autocorrelation peak intensity and the time difference position according to the autocorrelation coefficient.

[0053] Optionally, in some possible implementations of the present application, the extraction unit is specifically used to obtain a fluctuation range of a network transmission delay of the audio input signal within a collection time period;

[0054] The extraction unit is specifically configured to determine the detection duration according to the network transmission delay fluctuation range;

[0055] The extraction unit is specifically configured to determine the autocorrelation coefficient according to the repeatability of the audio input signal within the detection duration.

[0056] Optionally, in some possible implementations of the present application, the extraction unit is specifically used to process the audio input signal based on a preprocessing rule, and the preprocessing rule includes determining noise characteristics in at least one dimension;

[0057] The extraction unit is specifically configured to determine the autocorrelation coefficient according to the repeatability of the processed audio input signal within a detection time period.

[0058] Optionally, in some possible implementations of the present application, the extraction unit is specifically configured to perform time-frequency analysis on the audio input signal to obtain a time-frequency signal;

[0059] The extraction unit is specifically used to filter according to the peak characteristics of the time-frequency signal to obtain the spectrum peak point;

[0060] The extraction unit is specifically used to count the occurrence probability of the peak point of the spectrum to obtain the peak point frequency;

[0061] The extraction unit is specifically configured to extract the frequency spectrum feature according to the peak point frequency.

[0062] Optionally, in some possible implementations of the present application, the extraction unit is specifically used to obtain an energy ratio of the frequency point energy of each frequency point in the time-frequency signal to the total energy of the time-frequency signal according to the peak point frequency;

[0063] The extraction unit is specifically configured to determine the energy stationarity of the time-frequency signal according to the energy ratio to obtain the frequency spectrum feature.

[0064] Optionally, in some possible implementations of the present application, the extraction unit is specifically used to obtain a harmonic signal corresponding to the time-frequency signal;

[0065] The extraction unit is specifically configured to obtain the frequency energy of each frequency point in the time-frequency signal according to the peak point frequency;

[0066] The extraction unit is specifically configured to determine the frequency spectrum feature according to a ratio of the frequency point energy to the energy corresponding to the harmonic signal.

[0067] Optionally, in some possible implementations of the present application, the extraction unit is specifically used to determine a frequency point corresponding to the peak point frequency;

[0068] The extraction unit is specifically used to determine the frequency stability according to the fluctuation of adjacent frequency points among the frequency points corresponding to the peak point frequency, so as to obtain the frequency spectrum characteristics.

[0069] Optionally, in some possible implementations of the present application, the extraction unit is specifically configured to determine a candidate peak point according to a peak feature of the time-frequency signal;

[0070] The extraction unit is specifically used to select the candidate peak points that meet preset conditions to obtain spectrum peak points.

[0071] Optionally, in some possible implementations of the present application, the fusion unit is specifically used to determine a weighting parameter according to the periodic feature, and the weighting parameter is used to indicate the number of samples in a period in the periodic feature;

[0072] The fusion unit is specifically configured to assign weighting factors to the periodic feature and the spectrum feature according to the weighting parameter;

[0073] The fusion unit is specifically used to fuse the periodic feature after the weighting factor is assigned and the spectrum feature after the weighting factor is assigned to obtain the fusion feature.

[0074] Optionally, in some possible implementations of the present application, the fusion unit is specifically used to determine the network delay characteristics of the periodic characteristics and the spectrum characteristics according to the weighting parameter;

[0075] The fusion unit is specifically configured to assign weighting factors to the periodicity feature and the spectrum feature according to the network delay feature.

[0076] A third aspect of the present application provides a method for voice communication, including:

[0077] Get the audio input signal during the voice call;

[0078] Performing howling detection on the audio input signal to obtain a howling detection result, wherein the howling detection includes the howling detection method according to the first aspect or any one of the first aspects;

[0079] The corresponding call parameters during the voice call are adjusted according to the howling detection result.

[0080] A fourth aspect of the present application provides a voice call device, including:

[0081] An acquisition unit, used to acquire an audio input signal during a voice call;

[0082] a detection unit, configured to perform a howling detection on the audio input signal to obtain a howling detection result, wherein the howling detection includes the howling detection method according to the first aspect or any one of the first aspects;

[0083] The adjusting unit is used to adjust the corresponding call parameters during the voice call according to the howling detection result.

[0084] The fifth aspect of the present application provides a computer device, including: a memory, a processor and a bus system; the memory is used to store program code; the processor is used to execute the method for howling detection described in the first aspect or any one of the first aspects, or the method for voice call described in the third aspect according to the instructions in the program code.

[0085] In a sixth aspect, the present application provides a computer-readable storage medium, which stores instructions. When the computer-readable storage medium is executed on a computer, the computer executes the method for howling detection described in the first aspect or any one of the first aspects, or the method for voice communication described in the third aspect.

[0086] It can be seen from the above technical solutions that the embodiments of the present application have the following advantages:

[0087] By acquiring the audio input signal during the communication process, and extracting the periodic features and spectrum features of the audio input signal, wherein the periodic features include the autocorrelation peak intensity and the time difference position; then fusing the periodic features and the spectrum features to obtain the fused features; and then detecting the fused features based on at least one detection condition to obtain the howling detection result. Thus, the howling detection process based on the fused features is realized. Since the periodic features can be used to trace the time when the howling occurs in the early stage, and the spectrum features after the howling is formed are combined to make the final judgment, the probability of false detection is reduced and the accuracy of howling detection is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0088] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0089] Figure 1 A network architecture diagram for the operation of the howling detection system;

[0090] Figure 2 A flow chart of howling detection provided in an embodiment of the present application;

[0091] Figure 3 A flowchart of a method for howling detection provided in an embodiment of the present application;

[0092] Figure 4 A flowchart of a method for howling detection provided in an embodiment of the present application;

[0093] Figure 5 A flowchart of another method for howling detection provided in an embodiment of the present application;

[0094] Figure 6 A flowchart of another method for howling detection provided in an embodiment of the present application;

[0095] Figure 7 A schematic diagram of a howling detection scenario provided in an embodiment of the present application;

[0096] Figure 8 A flowchart of a voice call method provided in an embodiment of the present application;

[0097] Fig. 9 A schematic diagram of another scenario of howling detection provided in an embodiment of the present application;

[0098] Fig.10 A schematic diagram of another scenario of howling detection provided in an embodiment of the present application;

[0099] Fig.11 A schematic diagram of the structure of a howling detection device provided in an embodiment of the present application;

[0100] Fig.12 A schematic diagram of the structure of a voice communication device provided in an embodiment of the present application;

[0101] Fig.13 A schematic diagram of the structure of a terminal device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0102] The embodiment of the present application provides a method and related device for howling detection, which can be applied to a system or program in a terminal device that includes a howling detection function, by acquiring an audio input signal during the communication process, and extracting the periodic characteristics and spectrum characteristics of the audio input signal, wherein the periodic characteristics include the autocorrelation peak intensity and the time difference position; then the periodic characteristics and the spectrum characteristics are fused to obtain a fused characteristic; and then the fused characteristic is detected based on at least one detection condition to obtain a howling detection result. Thus, a howling detection process based on fused characteristics is realized. Since the periodic characteristics can be used to trace the time when the howling occurs in the early stage, and the spectrum characteristics after the howling is formed are combined to make a final judgment, the probability of false detection is reduced and the accuracy of howling detection is improved.

[0103] The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein, for example. In addition, the terms "including" and "corresponding to" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0104] First, some terms that may appear in the embodiments of the present application are explained.

[0105] Howling: The sound signal collected by the microphone is amplified by the speaker and then picked up by the microphone. The signal is continuously superimposed and amplified in the feedback loop, and the positive feedback produces an oscillation cycle, which leads to the phenomenon.

[0106] Howling point: The frequency point in the audio signal where the loop gain is greater than or equal to 1.

[0107] Short-time Fourier transform (STFT): A mathematical transform related to the Fourier transform that is used to determine the frequency and phase of a local sinusoidal wave of a time-varying signal.

[0108] Discrete cosine transform (DCT): Mainly used for data compression, it can convert spatial domain signals into frequency domain and has good decorrelation performance.

[0109] Modified discrete cosine transform (MDCT): A linear orthogonal lapped transform. It uses a time-domain aliasing cancellation technique, including a 50% time-domain overlapping window, to effectively overcome the edge effect in the windowed discrete cosine transform block processing operation without reducing the coding performance, thereby effectively removing the periodic noise generated by the edge effect.

[0110] Wavelet transform (WT) is a transform analysis method that can fully highlight the characteristics of certain aspects of the problem through transformation, can perform localized analysis of time and frequency, and gradually refine the signal at multiple scales through scaling and translation operations, ultimately achieving time subdivision at high frequencies and frequency subdivision at low frequencies. It can automatically adapt to the requirements of time-frequency signal analysis, thereby focusing on any details of the signal.

[0111] Harmonic: refers to the components greater than integer multiples of the fundamental frequency obtained by Fourier series decomposition of periodic non-sinusoidal alternating current, usually called high-order harmonics.

[0112] It should be understood that the howling detection method provided in the present application can be applied to a system or program including a howling detection function in a terminal device, such as a cloud conference platform. Specifically, the howling detection system can be run on Figure 1 In the network architecture shown in Figure 1 As shown in FIG. 1 , it is a network architecture diagram of the operation of the howling detection system. As can be seen from the figure, the howling detection system can provide howling detection with multiple information sources. The terminal establishes a connection with the server through the network to conduct a voice call process, and performs howling detection during the voice call to optimize the quality of the voice call. It can be understood that Figure 1 A variety of terminal devices are shown in FIG. 1 . In actual scenarios, more or fewer types of terminal devices may participate in the howling detection process. The specific number and type depend on the actual scenario and are not limited here. In addition, Figure 1One server is shown in the figure, but in actual scenarios, multiple servers may be involved, especially in scenarios where multiple content applications interact. The specific number of servers depends on the actual scenario.

[0113] In this embodiment, the server can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. The terminal can be a smart phone, a tablet computer, a laptop computer, a desktop computer, a smart speaker, a smart watch, a PSTN phone, etc., but is not limited to this. The terminal and the server can be directly or indirectly connected via wired or wireless communication, and this application does not limit this.

[0114] It should be noted that the howling detection method provided in this embodiment can also be performed offline, that is, without the participation of a server, and in this case the terminal is connected to other terminals locally to perform a howling detection process between the terminals.

[0115] It can be understood that the above-mentioned howling detection system can run on a personal mobile terminal, for example: as an application such as a cloud conferencing platform, it can also run on a server, and it can also run on a third-party device to provide howling detection to obtain the howling detection processing result of the information source; the specific howling detection system can be run in the above-mentioned device in the form of a program, or it can be run as a system component in the above-mentioned device, and it can also be used as a cloud service program. The specific operation mode depends on the actual scenario and is not limited here.

[0116] With the research and advancement of artificial intelligence technology, artificial intelligence technology has been studied and applied in many fields, such as common smart homes, smart wearable devices, virtual assistants, smart speakers, smart marketing, unmanned driving, automatic driving, drones, robots, smart medical care, smart customer service, etc. I believe that with the development of technology, artificial intelligence technology will be applied in more fields and play an increasingly important role.

[0117] For example, in audio communication, artificial intelligence can be used to detect and process howling. Howling is likely to occur when two communication devices are close to each other, or when two devices in different locations have echo leakage at the same time in long-distance communication, especially in cloud conference scenarios with multiple voices.

[0118] Cloud conferencing is an efficient, convenient and low-cost conferencing format based on cloud computing technology. Users only need to use the Internet interface and perform simple and easy-to-use operations to quickly and efficiently share voice, data files and videos with teams and customers around the world. The cloud conferencing service provider helps users operate complex technologies such as data transmission and processing during the meeting.

[0119] At present, domestic cloud conferences are mainly concentrated on service content based on the Software as a Service (SaaS) model, including telephone, network, video and other service forms. Video conferencing based on cloud computing is called cloud conferencing.

[0120] In the era of cloud conferencing, data transmission, processing, and storage are all handled by the computer resources of video conferencing manufacturers. Users no longer need to purchase expensive hardware or install cumbersome software. They only need to open a browser and log in to the corresponding interface to conduct efficient remote meetings.

[0121] The cloud conference system supports dynamic cluster deployment of multiple servers and provides multiple high-performance servers, which greatly improves the stability, security and availability of conferences. In recent years, video conferencing has been welcomed by many users because it can greatly improve communication efficiency, continuously reduce communication costs, and upgrade internal management levels. It has been widely used in various fields such as government, transportation, finance, operators, education, and enterprises. There is no doubt that after the use of cloud computing, video conferencing has a stronger appeal in terms of convenience, speed, and ease of use, which will surely stimulate the arrival of a new wave of video conferencing applications.

[0122] Generally, the howling detection process can start from the phenomenon of howling, that is, analyzing the spectrum characteristics of the howling sound. When the signal energy is concentrated in a narrow band, howling may exist, and howling detection is performed based on the spectrum characteristics of these narrow bands.

[0123] However, the process based on spectral features has limitations. For example, it is easy to misdetect other naturally existing single-frequency narrowband signals such as telephone ringing and alarms. In addition, the detection process can only detect severe howling after it occurs, which affects the accuracy and efficiency of howling detection.

[0124] In order to solve the above problems, the present application proposes a method for howling detection, which is applied to Figure 2 In the process framework of howling detection shown in FIG. Figure 2 As shown, it is a process architecture diagram of howling detection provided in an embodiment of the present application. First, the audio output signal emitted by the terminal is automatically collected, and then the periodic features and spectrum features in the audio output signal are extracted respectively, and the periodic features and the spectrum features are further fused to obtain fused features; thus, howling detection is performed according to the fused features.

[0125] It can be understood that the method provided in the present application can be written as a program as a processing logic in a hardware system, or as a howling detection device, which implements the above processing logic in an integrated or external manner. As an implementation method, the howling detection device obtains the audio input signal during the communication process, and extracts the periodic characteristics and spectrum characteristics of the audio input signal, wherein the periodic characteristics include the autocorrelation peak intensity and the time difference position; then fuses the periodic characteristics and the spectrum characteristics to obtain the fusion characteristics; and then detects the fusion characteristics based on at least one detection condition to obtain the howling detection result. Thus, the howling detection process based on the fusion characteristics is realized. Since the periodic characteristics can be used to trace the time when the early howling occurs, and the final judgment is made in combination with the spectrum characteristics after the howling is formed, the probability of false detection is reduced and the accuracy of howling detection is improved.

[0126] The solution provided in the embodiments of the present application involves technologies such as artificial intelligence speech processing, which is specifically described by the following embodiments:

[0127] Combined with the above process architecture, the method of howling detection in this application will be introduced below. Figure 3 , Figure 3 A flowchart of a method for howling detection provided in an embodiment of the present application, the embodiment of the present application at least includes the following steps:

[0128] 301. Obtain an audio input signal.

[0129] In this embodiment, the audio input signal is the output signal of the terminal device during the voice call, that is, the audio signal emitted by the microphone; because the terminal device collects the sound of the user's speech at the current moment through the microphone, it may also collect the sound of the remote caller played by the speaker of the terminal device, which may cause howling.

[0130] It is understandable that the process of acquiring the audio input signal may be real-time acquisition during a voice call, or may be voice sampling of a test voice during a test.

[0131] 302. Extract periodic features of the audio input signal.

[0132] In this embodiment, the periodic characteristics include autocorrelation peak intensity and time difference position; wherein the autocorrelation peak intensity is for finding the repetitive pattern of the howling, and the time difference position is for further determining the accuracy of the repetitive pattern.

[0133] Optionally, before determining the autocorrelation coefficient based on the repeatability of the audio input signal within the detection time, noise elimination can also be performed, that is, the preprocessing process, which specifically includes speech preprocessing such as echo elimination and noise reduction, so as to eliminate the interference of environmental noise and echo on the periodic feature extraction process and improve the accuracy of feature extraction.

[0134] Specifically, for the process of extracting the periodic characteristics of the audio input signal, first determine the autocorrelation coefficient according to the repeatability of the audio input signal within the detection time length; then determine the autocorrelation peak intensity and time difference position according to the autocorrelation coefficient. In one possible way, the autocorrelation coefficient can be calculated by the following formula:

[0135] C(n,m)=E[x(n)x(nm)]

[0136] Among them, x(n) represents a certain segment of speech, which can be the preprocessed speech mentioned above or the unprocessed one; C(n,m) represents the correlation strength between the input signal at time n and the time m before time n, that is, it measures whether the duration m is the repetition period of speech x.

[0137] Furthermore, the process of determining the autocorrelation peak strength and the time difference position according to the autocorrelation coefficient is to analyze the autocorrelation coefficient C(n,m) to extract the correlation peak T(n) and the period value m required for howling detection at time n. h (n). The specific calculation formula is as follows:

[0138]

[0139] Among them, M1 and M2 are the intervals for calculating periodic characteristics; C(n,m) represents the strength of the correlation between the input signal at time n and the time m before time n.

[0140] Optionally, the specific values ​​of M1 and M2 can be determined by the network transmission delay and sampling rate of the voice communication system; this is because the network transmission delay often fluctuates within a range, taking into account factors such as network quality and bandwidth load jitter. Specifically, first obtain the fluctuation range of the network transmission delay of the audio input signal within the acquisition time period; then determine the detection time according to the fluctuation range of the network transmission delay; and then determine the autocorrelation coefficient according to the repeatability of the audio input signal within the detection time period.

[0141] It is understandable that the fluctuation range of network transmission delay may be obtained through real-time testing or may be obtained through processing based on the fluctuation range in historical records.

[0142] In a possible scenario, assuming that the network transmission delay fluctuation range is [T1, T2], then the values ​​of M1 and M2 are:

[0143] M1=T1*f s

[0144] M2=T1*f s

[0145] Among them, T1 is the lower limit of one-way network transmission delay, T2 is the upper limit of two-way network transmission delay, and f s is the sampling rate, M1 and M2 are the intervals for periodic feature calculation.

[0146] By detecting the resulting network scene and extracting the periodic features of the audio input signal, interference caused by network fluctuations can be avoided, thereby improving the accuracy of periodic feature extraction.

[0147] 303. Extract frequency spectrum features of the audio input signal.

[0148] In this embodiment, the frequency spectrum feature is obtained based on the frequency spectrum distribution of the howling point, that is, the howling point generally appears in a frequency point indication area with local significance in the frequency spectrum diagram.

[0149] Specifically, in order to facilitate the extraction of spectral features, the audio input signal can be subjected to time-frequency analysis to obtain a time-frequency signal, that is, the audio input signal can be converted to an orthogonal transformation region. The audio input signal conversion process can be performed by STFT, DCT, MDCT, wavelet transform, etc. The specific method depends on the actual scenario and is not limited here.

[0150] Since the spectrum characteristics corresponding to howling generally fluctuate greatly, the response fluctuation can be referred to spectrum parameters such as peak frequency energy proportion, peak frequency harmonic proportion or peak frequency stability. The determination process of these parameters is explained below.

[0151] First, it is necessary to perform peak tracing on the signal after time-frequency analysis, screen out frequency points with significant local maximum characteristics as candidate peak points, and determine the spectrum peak based on the statistical probability of the candidate peak points appearing.

[0152] Optionally, in order to reduce the amount of data processing, frequency points with higher statistical probabilities among the candidate peak points may be selected; for example, 3-5 frequency points are selected and sent to subsequent modules. If the number of spectrum peak points is greater than 5, the first 5 with the highest statistical probabilities are selected.

[0153] Next, based on the statistical probabilities corresponding to these spectrum peak points, that is, the peak point frequencies, spectrum features of different dimensions are extracted.

[0154] (1) For the peak frequency energy ratio, first obtain the energy ratio of each frequency energy in the time-frequency signal to the total energy of the time-frequency signal according to the peak frequency; then determine the energy stability of the time-frequency signal according to the energy ratio to obtain the spectrum characteristics. Specifically, the following formula can be used for extraction:

[0155]

[0156] Where D(f,n) is the energy ratio of the peak frequency point, Q is the number of statistical frames, P is the frame length of each frame, and Y(f,n) is the ratio of the energy of the frequency point corresponding to the peak point frequency f at time n to the total energy, and is normalized to the log domain.

[0157] (2) For the peak frequency harmonic ratio, first obtain the harmonic signal corresponding to the time-frequency signal, for example, perform a third harmonic transformation; then obtain the frequency energy of each frequency point in the time-frequency signal according to the peak frequency; and determine the spectrum characteristics according to the ratio of the frequency energy to the energy corresponding to the harmonic signal. Specifically, the following formula can be used for extraction:

[0158]

[0159] Among them, X(f,n) is the energy corresponding to the frequency point at time n calculated based on the peak frequency point f, M is the harmonic order constant, g is the current harmonic order, and H(f,n) is the harmonic proportion of the peak frequency point.

[0160] (3) For the stability of the peak frequency, first determine the frequency point corresponding to the peak frequency; then determine the frequency stability based on the fluctuation of the adjacent frequency points in the frequency point corresponding to the peak frequency, thereby obtaining the spectrum characteristics. Specifically, the following formula can be used for extraction:

[0161] P(f(n),n)=s*P(f(n-1),n-1)+(1-s)[f(n)-f(n-1)]

[0162] Where s is the smoothing factor, f(n) is the peak frequency, and P(f,n) is the frequency smoothness corresponding to the frequency point at time n based on the peak frequency calculation.

[0163] It is understandable that for the spectrum parameters of different dimensions mentioned above, one or more of them can be used to obtain the spectrum characteristics; the specific form depends on the actual scenario. Specifically, when determining the spectrum characteristics based on the above three dimensions, the following formula can be used:

[0164]

[0165] Where L is the number of spectrum peak points, f iis the frequency value of each peak point, D(f,n) is the peak frequency energy ratio, H(f,n) is the peak frequency harmonic ratio, and P(f,n) is the stability based on the peak frequency; a1, a2, and a3 are weight parameters, which can be set by relevant personnel based on experience, or the terminal can automatically generate them based on historical records, for example, a1 = -0.5, a2 = 1, and a3 = -0.1. Among them, a1 is mainly used to eliminate false detections caused by music. For example, if it is used in a karaoke environment, a1 can be appropriately increased; a2 is mainly used for call howling detection, that is, a2 is appropriately increased in conference scenarios; a3 is mainly used to eliminate false detections of ringtones. If the scene is noisy, a3 can be appropriately increased. The specific value depends on the actual scene and is not limited here.

[0166] 304. Fuse the periodic feature and the spectrum feature to obtain a fused feature.

[0167] In this embodiment, different network delays may cause howling with different characteristics. For example, short network delays are more likely to cause howling based on spectrum characteristics, and long delays are more likely to cause howling based on periodic characteristics. Therefore, the fusion of periodic characteristics and spectrum characteristics can be performed based on weighting factors, that is, the weighting parameters are determined based on the periodic characteristics, and the weighting parameters are determined based on the m h (n) / f s The number of samples in the periodic feature is obtained; then, a weighting factor is assigned to the periodic feature and the spectrum feature according to the weighting parameter, wherein the weighting factor is set based on the network delay feature and can be assigned according to the relative proportion of the network delay feature; thus, the periodic feature after the weighting factor is assigned and the spectrum feature after the weighting factor is assigned are fused to obtain the spectrum feature. For details, refer to the following formula:

[0168]

[0169] Among them, W(n) is the fusion feature; T(n) is the periodic feature; F(n) is the spectrum feature; T1 is the lower limit of the one-way network transmission delay; T2 is the upper limit of the two-way network transmission delay; m h (n) is the period size value; f s is the sampling rate.

[0170] 305. Detect the fused feature based on at least one detection condition to obtain a howling detection result.

[0171] In this embodiment, the detection condition is set based on the probability or intensity of the howling point. That is, the fusion feature W(n) and the howling threshold W Thres For comparison, the formula is as follows:

[0172]

[0173] Among them, W(n) is the fusion feature, j is the detection time, and W Thres is the howling threshold, and Q is the number of statistical frames.

[0174] Furthermore, by judging that the statistical probability K(n) is greater than K Thres If the howling occurs, then the howling occurs, otherwise it is determined that the howling does not occur, and the corresponding time n is recorded.

[0175] In combination with the above embodiments, it can be seen that by acquiring the audio input signal in the communication process, and extracting the periodic features and spectrum features of the audio input signal, wherein the periodic features include the autocorrelation peak intensity and the time difference position; then the periodic features and the spectrum features are fused to obtain the fused features; and then the fused features are detected based on at least one detection condition to obtain the howling detection result. Thus, the howling detection process based on the fused features is realized. Since the periodic features can be used to trace the time when the early howling occurs, and the spectrum features after the howling is formed are combined to make the final judgment, the probability of false detection is reduced and the accuracy of howling detection is improved.

[0176] The above embodiment introduces the process of howling detection. The following describes the processing process of the audio input signal in combination with the specific process distribution. Figure 4 As shown, it is a flow chart of a method for howling detection provided by an embodiment of the present application; in the figure, the audio input signal first passes through the speech preprocessing module 401, and then the preprocessed speech is input into the periodic feature module 402 to extract the periodic feature. In addition, the audio input signal is input into the spectrum feature module 403 in another thread to extract the spectrum feature, and then the periodic feature and the spectrum feature are input into the fusion feature module 404 to obtain the fusion feature, and the howling detection process is performed based on the feature.

[0177] Specifically, for the periodic feature module 402, such as Figure 5 As shown, it is a flow chart of another method for howling detection provided by the embodiment of the present application; the process includes two parts: speech autocorrelation calculation module 501 and correlation feature extraction module 502. The specific process can be referred to Figure 3 The relevant description of step 302 in the embodiment is not repeated here.

[0178] Specifically, for the spectrum feature module 403, Figure 6As shown, it is a process framework diagram of another howling detection method provided by an embodiment of the present application; first, the audio input signal is input into the time-frequency analysis module 601, and then into the spectrum peak tracing module 602, and then according to the obtained spectrum peak, the spectrum parameters of at least one module among the peak frequency energy proportion module 603, the peak harmonic energy ratio module 604 or the peak frequency smoothness module 605 are obtained, and combined in the howling spectrum feature extraction module 606 to obtain the spectrum feature.

[0179] It can be understood that the above module connection relationship is a logical schematic, which can be a virtual process association or a physical device association, which is not limited here.

[0180] The above embodiment introduces the process of howling detection. In another possible scenario, Figure 7 , is a schematic diagram of a scenario of audio signal processing provided by an embodiment of the present application, which shows a loop in a terminal external speaker scenario. When the two terminals are close to each other, the sound comes out of the speaker of the right terminal and is picked up by the microphone of the left terminal. After pre-processing and signal conversion, it is sent to the right terminal through the network. It is played through the speaker and then picked up by the microphone of the left terminal. This cycle continues. If the loop gain is greater than or equal to 1 at a certain frequency point and the phase is positive, then this point will form a howling point.

[0181] Next, we will use game applications as a specific scenario. Figure 7 For an introduction to the applications in Figure 8 , Figure 8 A flowchart of another method for howling detection provided in an embodiment of the present application, the embodiment of the present application at least includes the following steps:

[0182] 801. Acquire an audio input signal during a voice call.

[0183] 802. Perform howling detection on the audio input signal to obtain a howling detection result.

[0184] In this embodiment, the relevant features of steps 801 and 802 are Figure 3 The steps 301-305 shown are similar and will not be described in detail here.

[0185] 803. Adjust corresponding call parameters during the voice call according to the howling detection result.

[0186] In this embodiment, the call parameter adjustment may be to detect howling and alert the user, or to detect howling and turn off the microphone, or to detect howling and turn off the speaker, or to detect howling and lower the volume of the microphone or speaker.

[0187] Specifically, Fig. 9 As shown, it is a schematic diagram of another scenario of howling detection provided in an embodiment of the present application. The figure shows a scenario of a game voice call. When howling is detected, the call feature A1 can provide a reminder indication, for example, in the form of flashing or shaking.

[0188] In another scenario, Fig.10 As shown, it is a schematic diagram of another scenario of howling detection provided in an embodiment of the present application. The figure shows a cloud conference scenario. When howling is detected, a window prompt will pop up to indicate that howling has been detected and associated processing operations, for example: turning off the microphone.

[0189] In combination with the above embodiments, it can be seen that by automatically performing a background process for howling detection and timely performing voice adjustment or reminders during a voice call, the degradation of call quality due to howling is avoided, the quality of the user's voice call is improved, and the user experience is enhanced.

[0190] In order to better implement the above solution of the embodiment of the present application, the following also provides related devices for implementing the above solution. Fig.11 , Fig.11 This is a schematic diagram of the structure of a howling detection device provided in an embodiment of the present application. The howling detection device 1100 includes:

[0191] An acquisition unit 1101 is used to acquire an audio input signal;

[0192] An extraction unit 1102, configured to extract periodic features of the audio input signal, wherein the periodic features include autocorrelation peak intensity and time difference position;

[0193] The extraction unit 1102 is further configured to extract a spectrum feature of the audio input signal, where the spectrum feature is obtained by fusing spectrum parameters of at least one dimension, and the spectrum parameter is related to the spectrum distribution of the howling point;

[0194] A fusion unit 1103, configured to fuse the periodic feature and the spectrum feature to obtain a fusion feature;

[0195] The detection unit 1104 is configured to detect the fused feature based on at least one detection condition to obtain a howling detection result, wherein the detection condition is set based on the occurrence probability or intensity of the howling point.

[0196] Optionally, in some possible implementations of the present application, the extraction unit 1102 is specifically configured to determine an autocorrelation coefficient according to the repeatability of the audio input signal within a detection duration;

[0197] The extraction unit 1102 is specifically configured to determine the autocorrelation peak intensity and the time difference position according to the autocorrelation coefficient.

[0198] Optionally, in some possible implementations of the present application, the extraction unit 1102 is specifically used to obtain a fluctuation range of a network transmission delay of the audio input signal within a collection time period;

[0199] The extraction unit 1102 is specifically configured to determine the detection duration according to the network transmission delay fluctuation range;

[0200] The extraction unit 1102 is specifically configured to determine an autocorrelation coefficient according to the repeatability of the audio input signal within the detection duration.

[0201] Optionally, in some possible implementations of the present application, the extraction unit 1102 is specifically configured to process the audio input signal based on a preprocessing rule, wherein the preprocessing rule includes determining noise features in at least one dimension;

[0202] The extraction unit 1102 is specifically configured to determine the autocorrelation coefficient according to the repeatability of the processed audio input signal within a detection time period.

[0203] Optionally, in some possible implementations of the present application, the extraction unit 1102 is specifically configured to perform time-frequency analysis on the audio input signal to obtain a time-frequency signal;

[0204] The extraction unit 1102 is specifically used to filter according to the peak characteristics of the time-frequency signal to obtain the spectrum peak point;

[0205] The extraction unit 1102 is specifically used to count the occurrence probability of the spectrum peak point to obtain the peak point frequency;

[0206] The extraction unit 1102 is specifically configured to extract the frequency spectrum feature according to the peak point frequency.

[0207] Optionally, in some possible implementations of the present application, the extraction unit 1102 is specifically configured to obtain, according to the peak point frequency, an energy ratio of the frequency point energy of each frequency point in the time-frequency signal to the total energy of the time-frequency signal;

[0208] The extraction unit 1102 is specifically configured to determine the energy stationarity of the time-frequency signal according to the energy ratio to obtain the frequency spectrum feature.

[0209] Optionally, in some possible implementations of the present application, the extraction unit 1102 is specifically used to obtain a harmonic signal corresponding to the time-frequency signal;

[0210] The extraction unit 1102 is specifically configured to obtain the frequency energy of each frequency point in the time-frequency signal according to the peak point frequency;

[0211] The extraction unit 1102 is specifically configured to determine the frequency spectrum feature according to a ratio of the frequency point energy to the energy corresponding to the harmonic signal.

[0212] Optionally, in some possible implementations of the present application, the extraction unit 1102 is specifically configured to determine a frequency point corresponding to the peak point frequency;

[0213] The extraction unit 1102 is specifically configured to determine the frequency stability according to the fluctuation of adjacent frequency points among the frequency points corresponding to the peak point frequency, so as to obtain the frequency spectrum feature.

[0214] Optionally, in some possible implementations of the present application, the extraction unit 1102 is specifically configured to determine a candidate peak point according to a peak feature of the time-frequency signal;

[0215] The extraction unit 1102 is specifically configured to select the candidate peak points that meet preset conditions to obtain spectrum peak points.

[0216] Optionally, in some possible implementations of the present application, the fusion unit 1103 is specifically used to determine a weighting parameter according to the periodic feature, and the weighting parameter is used to indicate the number of samples in a period in the periodic feature;

[0217] The fusion unit 1103 is specifically configured to assign a weighting factor to the periodic feature and the spectrum feature according to the weighting parameter;

[0218] The fusion unit 1103 is specifically used to fuse the periodic feature after the weighting factor is assigned and the spectrum feature after the weighting factor is assigned to obtain the fused feature.

[0219] Optionally, in some possible implementations of the present application, the fusion unit 1103 is specifically configured to determine a network delay feature of the periodic feature and the spectrum feature according to the weighted parameter;

[0220] The fusion unit 1103 is specifically configured to assign weighting factors to the periodicity feature and the spectrum feature according to the network delay feature.

[0221] By acquiring the audio input signal during the communication process, and extracting the periodic features and spectrum features of the audio input signal, wherein the periodic features include the autocorrelation peak intensity and the time difference position; then fusing the periodic features and the spectrum features to obtain the fused features; and then detecting the fused features based on at least one detection condition to obtain the howling detection result. Thus, the howling detection process based on the fused features is realized. Since the periodic features can be used to trace the time when the howling occurs in the early stage, and the spectrum features after the howling is formed are combined to make the final judgment, the probability of false detection is reduced and the accuracy of howling detection is improved.

[0222] The present application also provides a voice communication device 1200. Fig.12 FIG. 1 is a schematic diagram of a structure of a voice call provided by an embodiment of the present application, which specifically includes:

[0223] The acquisition unit 1201 is used to acquire an audio input signal during a voice call;

[0224] A detection unit 1202, configured to perform a howling detection on the audio input signal to obtain a howling detection result, wherein the howling detection includes the howling detection method according to the first aspect or any one of the first aspect;

[0225] The adjusting unit 1203 is configured to adjust corresponding call parameters during the voice call according to the howling detection result.

[0226] The present application also provides a terminal device, such as Fig.13 , is a schematic diagram of the structure of another terminal device provided by the embodiment of the present application. For the sake of convenience, only the part related to the embodiment of the present application is shown. For specific technical details not disclosed, please refer to the method part of the embodiment of the present application. The terminal can be any terminal device including a mobile phone, a tablet computer, a personal digital assistant (PDA), a point of sales (POS), a car computer, etc.

[0227] In the embodiment of the present application, the processor 1380 included in the terminal also has the function of executing each step of the above-mentioned page processing method.

[0228] In an embodiment of the present application, a computer-readable storage medium is also provided. The computer-readable storage medium stores a howling detection instruction. When the computer-readable storage medium is run on a computer, the computer executes the above-mentioned Figures 3 to 10 The illustrated embodiment describes the steps performed by the howling detection device in the method.

[0229] The present application also provides a computer program product including a howling detection instruction, which, when executed on a computer, enables the computer to execute the above-mentioned Figures 3 to 10 The illustrated embodiment describes the steps performed by the howling detection device in the method.

[0230] The embodiment of the present application also provides a howling detection system, which may include Fig.11 The howling detection device in the described embodiment, or Fig.12 The voice communication device in the described embodiment, or Fig.13 The terminal device described.

[0231] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0232] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for howling detection, characterized in that: include: Get audio input signal; Extracting periodic features of the audio input signal, wherein the periodic features include autocorrelation peak intensity and time difference position; The time difference position is a period value required for howling detection at time n extracted according to the autocorrelation coefficient; the autocorrelation coefficient indicates the strength of the correlation between the input signal at time n and the time m before time n; Extracting a frequency spectrum feature of the audio input signal, where the frequency spectrum feature is obtained by fusing frequency spectrum parameters of at least one dimension, and the frequency spectrum parameters are related to the frequency spectrum distribution of the howling point; fusing the periodic feature and the spectrum feature to obtain a fused feature; The fusion feature is detected based on at least one detection condition to obtain a howling detection result, wherein the detection condition is set based on the occurrence probability or intensity of the howling point.

2. The method according to claim 1, characterized in that The extracting the periodic feature of the audio input signal comprises: Determining an autocorrelation coefficient according to the repeatability of the audio input signal within a detection time period; The autocorrelation peak intensity and the time difference position are determined according to the autocorrelation coefficient.

3. The method according to claim 2, characterized in that The step of determining the autocorrelation coefficient according to the repeatability of the audio input signal within the detection time period includes: Obtaining a network transmission delay fluctuation range of the audio input signal within a collection time period; Determining the detection duration according to the fluctuation range of the network transmission delay; An autocorrelation coefficient is determined according to the repeatability of the audio input signal within the detection time period.

4. The method according to claim 2, characterized in that: The step of determining the autocorrelation coefficient according to the repeatability of the audio input signal within the detection time period includes: Processing the audio input signal based on a preprocessing rule, the preprocessing rule comprising determining noise characteristics in at least one dimension; The autocorrelation coefficient is determined according to the repeatability of the processed audio input signal within the detection time period.

5. The method according to claim 1, characterized in that The extracting the frequency spectrum feature of the audio input signal comprises: Performing time-frequency analysis on the audio input signal to obtain a time-frequency signal; Screening is performed according to the peak characteristics of the time-frequency signal to obtain a spectrum peak point; Counting the occurrence probability of the peak point of the spectrum to obtain the peak point frequency; The frequency spectrum feature is extracted according to the peak point frequency.

6. The method according to claim 5, characterized in that The extracting the frequency spectrum feature according to the peak point frequency includes: Acquire, according to the peak point frequency, an energy ratio of the frequency point energy of each frequency point in the time-frequency signal to the total energy of the time-frequency signal; The energy stationarity of the time-frequency signal is determined according to the energy ratio to obtain the frequency spectrum feature.

7. The method according to claim 5, characterized in that The extracting the frequency spectrum feature according to the peak point frequency includes: Obtaining a harmonic signal corresponding to the time-frequency signal; Acquire the frequency energy of each frequency point in the time-frequency signal according to the peak point frequency; The frequency spectrum feature is determined according to the ratio of the frequency point energy to the energy corresponding to the harmonic signal.

8. The method according to claim 5, characterized in that The extracting the frequency spectrum feature according to the peak point frequency includes: Determine the frequency point corresponding to the peak point frequency; The frequency stability is determined according to the fluctuation of adjacent frequency points among the frequency points corresponding to the peak point frequency, so as to obtain the frequency spectrum characteristics.

9. The method according to claim 5, characterized in that The screening according to the peak characteristics of the time-frequency signal to obtain the spectrum peak point includes: Determining candidate peak points according to the peak characteristics of the time-frequency signal; The candidate peak points satisfying the preset conditions are selected to obtain the spectrum peak points.

10. The method according to claim 1, characterized in that The fusing the periodic feature and the spectrum feature to obtain a fused feature includes: Determining a weighting parameter according to the periodic feature, the weighting parameter being used to indicate the number of samples within a period in the periodic feature; assigning weighting factors to the periodicity characteristic and the spectrum characteristic according to the weighting parameter; The periodic feature after the weighting factor is assigned and the spectrum feature after the weighting factor is assigned are fused to obtain the fused feature.

11. The method according to claim 10, characterized in that The allocating weighting factors to the periodicity feature and the spectrum feature according to the weighting parameter comprises: Determine the network delay characteristics of the periodic characteristics and the spectrum characteristics according to the weighting parameters; A weighting factor is assigned to the periodicity characteristic and the spectrum characteristic according to the network delay characteristic.

12. A method for voice communication, characterized in that: include: Get the audio input signal during the voice call; Performing howling detection on the audio input signal to obtain a howling detection result, wherein the howling detection includes the howling detection method according to any one of claims 1 to 11; The corresponding call parameters during the voice call are adjusted according to the howling detection result.

13. A device for howling detection, characterized in that: include: An acquisition unit, used for acquiring an audio input signal; An extraction unit, used to extract periodic features of the audio input signal, wherein the periodic features include autocorrelation peak intensity and time difference position; The time difference position is a period value required for howling detection at time n extracted according to the autocorrelation coefficient; the autocorrelation coefficient indicates the strength of the correlation between the input signal at time n and the time m before time n; The extraction unit is further used to extract a spectrum feature of the audio input signal, where the spectrum feature is obtained by fusing spectrum parameters of at least one dimension, and the spectrum parameter is related to the spectrum distribution of the howling point; A fusion unit, used for fusing the periodic feature and the spectrum feature to obtain a fusion feature; A detection unit is used to detect the fusion feature based on at least one detection condition to obtain a howling detection result, wherein the detection condition is set based on the occurrence probability or intensity of the howling point.

14. The device according to claim 13, characterized in that The extraction unit is specifically used for: Determining an autocorrelation coefficient according to the repeatability of the audio input signal within a detection time period; The autocorrelation peak intensity and the time difference position are determined according to the autocorrelation coefficient.

15. The device according to claim 14, characterized in that The extraction unit is specifically used for: Obtaining a network transmission delay fluctuation range of the audio input signal within a collection time period; Determining the detection duration according to the fluctuation range of the network transmission delay; An autocorrelation coefficient is determined according to the repeatability of the audio input signal within the detection time period.

16. The device according to claim 14, characterized in that The extraction unit is specifically used for: Processing the audio input signal based on a preprocessing rule, the preprocessing rule comprising determining noise characteristics in at least one dimension; The autocorrelation coefficient is determined according to the repeatability of the processed audio input signal within the detection time period.

17. The device according to claim 13, characterized in that The extraction unit is specifically used for: Performing time-frequency analysis on the audio input signal to obtain a time-frequency signal; Screening is performed according to the peak characteristics of the time-frequency signal to obtain a spectrum peak point; Counting the occurrence probability of the peak point of the spectrum to obtain the peak point frequency; The frequency spectrum feature is extracted according to the peak point frequency.

18. The device according to claim 17, characterized in that The extraction unit is specifically used for: Acquire, according to the peak point frequency, an energy ratio of the frequency point energy of each frequency point in the time-frequency signal to the total energy of the time-frequency signal; The energy stationarity of the time-frequency signal is determined according to the energy ratio to obtain the frequency spectrum feature.

19. The device according to claim 17, characterized in that The extraction unit is specifically used for: Obtaining a harmonic signal corresponding to the time-frequency signal; Acquire the frequency energy of each frequency point in the time-frequency signal according to the peak point frequency; The frequency spectrum feature is determined according to the ratio of the frequency point energy to the energy corresponding to the harmonic signal.

20. The device according to claim 17, characterized in that The extraction unit is specifically used for: Determine the frequency point corresponding to the peak point frequency; The frequency stability is determined according to the fluctuation of adjacent frequency points among the frequency points corresponding to the peak point frequency, so as to obtain the frequency spectrum characteristics.

21. The device according to claim 17, characterized in that The extraction unit is specifically used for: Determining candidate peak points according to the peak characteristics of the time-frequency signal; The candidate peak points satisfying the preset conditions are selected to obtain the spectrum peak points.

22. The device according to claim 13, characterized in that The fusion unit is specifically used for: Determining a weighting parameter according to the periodic feature, the weighting parameter being used to indicate the number of samples within a period in the periodic feature; assigning weighting factors to the periodicity characteristic and the spectrum characteristic according to the weighting parameter; The periodic feature after the weighting factor is assigned and the spectrum feature after the weighting factor is assigned are fused to obtain the fused feature.

23. The device according to claim 22, characterized in that The fusion unit is specifically used for: Determine the network delay characteristics of the periodic characteristics and the spectrum characteristics according to the weighting parameters; A weighting factor is assigned to the periodicity characteristic and the spectrum characteristic according to the network delay characteristic.

24. A device for voice communication, characterized in that: include: An acquisition unit, used to acquire an audio input signal during a voice call; a detection unit, configured to perform a howling detection on the audio input signal to obtain a howling detection result, wherein the howling detection comprises the howling detection method according to any one of claims 1 to 11; The adjusting unit is used to adjust the corresponding call parameters during the voice call according to the howling detection result.

25. A computer device, characterized in that: The computer device comprises a processor and a memory: The memory is used to store program codes; the processor is used to execute the howling detection method according to any one of claims 1 to 11, or the voice call method according to claim 12 according to the instructions in the program code.

26. A computer-readable storage medium storing instructions, which, when executed on a computer, enable the computer to execute the method for howling detection as described in any one of claims 1 to 11, or the method for voice communication as described in claim 12.

27. A computer program product comprising a howling detection instruction, characterized in that: When it is run on a computer, the computer is enabled to execute the howling detection method described in any one of claims 1 to 11, or the voice communication method described in claim 12.

Images