Active noise reduction system and abnormal sound detection method and device thereof
By using the target frequency resolution and Fourier transform in the active noise reduction system, combined with spectrum analysis and A-weighting processing, the problem of not being able to identify stable single-frequency antonyms and large calculations in the prior art is solved, and high efficiency is achieved singular sound detection.
Patent Information
- Application Number
- CN202411080535.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-08
- Publication Date
- 2024-11-12
- Estimated Expiration
- 2044-08-08
AI Technical Summary
The existing active noise reduction system is prone to howling when the feedback signal is greater than the noise signal, and cannot effectively identify stable single-frequency antonyms, and the calculation amount is large, which cannot meet the timeliness of antonyms detection.
The frequency resolution is used to perform Fourier transform on the noise signal, and the frequency resolution is determined by finding the optimization to accurately identify the different sounds, and determine whether there is a stable single-frequency different sound by the total amplitude of the spectrum power and the difference threshold. Combined with the A-meter Weight processing and segmentation strategies improve detection accuracy and rate.
It improves the accuracy and speed of abnormal sound detection, meets the timeliness requirements, can accurately identify stable single-frequency abnormal sounds, reduces the calculation amount, and avoids misjudgment of howling phenomena.
Smart Images

Figure CN118609535B_ABST
Abstract
Description
Technical Field
[0001] The present disclosure relates to the technical field of active noise reduction, and in particular to an active noise reduction system and an abnormal sound detection method and device thereof. Background Art
[0002] In the active noise reduction system, the noise reduction system uses the superposition interference effect of the control signal (anti-phase noise) generated according to the input signal (including the collected noise signal and the feedback signal) and the noise signal to achieve the purpose of noise reduction by reducing the noise signal. However, in this process, when the feedback signal is greater than the noise signal, a howling phenomenon will occur. Even if the feedback signal is less than the noise signal, when the ratio of the feedback signal to the noise signal is greater than a threshold, some stable single-frequency abnormal sounds will still be emitted. The amplitude of these stable single-frequency abnormal sounds is much smaller than the noise during howling, and there is no obvious difference between the amplitude of the speaker control signal generated when there is a stable single-frequency abnormal sound and the amplitude of the speaker control signal generated when the system is working normally. Therefore, through the traditional howling monitoring scheme based on the amplitude of the speaker control signal in the prior art, it is impossible to effectively identify this stable single-frequency abnormal sound phenomenon, and the large amount of calculation cannot meet the timeliness requirements of abnormal sound detection. Summary of the invention
[0003] The technical problem to be solved by the present disclosure is to overcome the above-mentioned defects in the prior art and to provide an active noise reduction system and an abnormal sound detection method and device thereof.
[0004] The present invention solves the above technical problems through the following technical solutions:
[0005] In a first aspect, a method for detecting an abnormal sound is provided, which is applied to an active noise reduction system, wherein the active noise reduction system includes a microphone and a speaker; the method for detecting an abnormal sound includes:
[0006] Acquire a first noise signal during the operation of the active noise reduction system; the first noise signal includes a noise signal of a noise source collected by the microphone and / or a control signal of the speaker;
[0007] The first noise signal is subjected to Fourier transform at a target frequency resolution, and abnormal sound detection is performed on the active noise reduction system according to the Fourier transform result; wherein the target frequency resolution is obtained by optimization, and the optimization goal is to accurately identify the abnormal sound and minimize the calculation amount of the abnormal sound detection.
[0008] Optionally, the step of determining the target frequency resolution includes:
[0009] Acquire an abnormal sample; the abnormal sample is a noise signal containing an abnormal sound;
[0010] The frequency resolution is updated iteratively, and the frequency resolution corresponding to when the abnormal sound of the abnormal sample is detected and the calculation amount is minimized is determined as the target frequency resolution.
[0011] Optionally, obtaining a first noise signal during operation of the active noise reduction system includes:
[0012] A first noise signal during the operation of the active noise reduction system is obtained at a target sampling frequency; the target sampling frequency is determined according to an abnormal sound frequency and / or a computing power of an analysis device of the first noise signal; and the abnormal sound frequency is determined according to a noise sample containing an abnormal sound.
[0013] Optionally, the target sampling frequency is greater than or equal to twice the frequency of the abnormal sound;
[0014] And / or, the computing power is characterized by the number of frequency domain data of the first noise signal analyzed by the analysis device per unit time; and the target sampling frequency is less than or equal to the number.
[0015] Optionally, acquiring a first noise signal during operation of the active noise reduction system at a target sampling frequency includes:
[0016] An original noise signal is acquired, and the original noise signal is resampled at the target sampling frequency to obtain the first noise signal.
[0017] Optionally, performing abnormal sound detection on the active noise reduction system according to the Fourier transform result includes:
[0018] Determine the total amplitude of the spectral power of the first noise signal according to the Fourier transform result;
[0019] In response to the total amplitude of the spectrum power being within a preset power range, it is determined that the active noise reduction system has a stable single-frequency abnormal sound; in response to the difference between the total amplitude of the spectrum power and the noise peak being less than a difference threshold, it is determined that the active noise reduction system has a stable single-frequency abnormal sound; wherein the lower limit of the preset power range is greater than the maximum power of the noise signal of the noise source, and the upper limit of the preset power range is less than the power of the first noise signal when howling occurs in the active noise reduction system;
[0020] and / or, in response to the difference between the total amplitude of the spectrum power and the noise peak being less than a first difference threshold, determining that a stable single-frequency abnormal sound exists in the active noise reduction system;
[0021] And / or, in response to the presence of a noise segment in the first noise signal whose power ratio is greater than or equal to a ratio threshold, determining that a stable single-frequency abnormal sound exists in the active noise reduction system; the power ratio is the ratio of the power of the noise segment to the total amplitude of the spectrum power;
[0022] And / or, in response to the difference between the total amplitude of the spectrum power and the filter power being greater than or equal to a second difference threshold, determining that the active noise reduction system has a stable single-frequency abnormal sound; the filter power is the result of filtering out the noise peak from the total amplitude of the spectrum power;
[0023] And / or, in response to the presence of a signal segment having a power difference with an adjacent signal segment greater than a third difference threshold, it is determined that a stable single-frequency alien sound exists.
[0024] Optionally, the preset power range is determined by:
[0025] Acquire abnormal samples and normal samples; the abnormal samples are noise signals containing abnormal sounds, and the normal samples are noise signals not containing abnormal sounds;
[0026] Constructing an objective function; the objective function is constructed according to at least one of the following parameters: the number of missed detections of abnormal samples, the number of false detections of normal samples, the percentage of missed detections of abnormal samples, and the percentage of false detections of normal samples;
[0027] Optimizing the objective function within a first preset optimization range of the upper limit value and / or the lower limit value to determine a final preset power range;
[0028] And / or, the difference threshold is determined by:
[0029] Acquire abnormal samples and normal samples; the abnormal samples are noise signals containing abnormal sounds, and the normal samples are noise signals not containing abnormal sounds;
[0030] Constructing an objective function; the objective function is constructed according to at least one of the following parameters: the number of missed detections of abnormal samples, the number of false detections of normal samples, the percentage of missed detections of abnormal samples, and the percentage of false detections of normal samples;
[0031] The objective function is optimized within a second preset optimization range of the difference threshold to determine a final difference threshold.
[0032] Optionally, the first preset optimization range is [74dB, 91dB];
[0033] And / or, the second preset optimization range is [1dB, 10dB];
[0034] And / or, the optimization algorithm includes at least one of the following algorithms: genetic algorithm, simulated annealing algorithm, particle swarm algorithm, Bayesian optimization algorithm, hill climbing algorithm;
[0035] And / or, the objective function is the ratio of the number of anomalies to the total number of samples, and the number of anomalies is the sum of the number of missed detections of the abnormal samples and the number of false detections of normal samples.
[0036] Optionally, determining the total amplitude of the spectral power of the first noise signal according to the Fourier transform result includes: performing A-weighting processing on the Fourier transform result, and determining the total amplitude of the spectral power of the first noise signal according to the A-weighting processing result;
[0037] and / or, dividing the Fourier transform result into at least two noise segments, and determining the total amplitude of the spectrum power according to the integration of each noise segment;
[0038] And / or, in response to the total amplitude of the spectrum power being within a preset power range, determining that the active noise reduction system has a stable single-frequency abnormal sound, including: in response to the total amplitude of the spectrum power being within the preset power range, the number of occurrences being greater than a number threshold, determining that the active noise reduction system has a stable single-frequency abnormal sound.
[0039] Optionally, before obtaining the first noise signal during the operation of the active noise reduction system, the method further includes:
[0040] When the active noise reduction system is not started, obtaining a second noise signal from a noise source;
[0041] Performing segmentation processing on the second noise signal so that the power proportion of each noise segment of the second noise signal is less than a proportion threshold, and determining a segmentation strategy for the second noise signal;
[0042] Performing Fourier transform on the first noise signal includes:
[0043] The first noise signal is processed in segments according to the segmentation strategy, and Fourier transform is performed on the noise segments obtained by segmentation.
[0044] Optionally, before obtaining the first noise signal during the operation of the active noise reduction system, the method further includes:
[0045] When the active noise reduction system is not started, obtaining a second noise signal from a noise source;
[0046] performing octave calculation on the second noise signal to determine a segmentation strategy matching the second noise signal;
[0047] Segmenting the second noise signal according to the segmentation strategy, and calculating the characteristic value of each noise segment of the second noise signal; the characteristic value includes power and / or decibel value;
[0048] Determine a target gain of a first target noise segment whose characteristic value is greater than a characteristic value threshold; after filtering the first target noise segment with a bandpass filter of the target gain, the characteristic value of the first target noise segment is less than or equal to the characteristic value threshold;
[0049] Performing Fourier transform on the first noise signal includes:
[0050] The first noise signal is segmented to obtain at least two noise segments, and a second target noise segment contained in the first noise signal is filtered using a bandpass filter with a target gain, and an unfiltered noise segment and a filtered noise segment in the first noise signal are respectively Fourier transformed; wherein a frequency band of the second target noise segment includes a frequency band of the first target noise segment.
[0051] In a second aspect, a control method for an active noise reduction system is provided, comprising:
[0052] Performing abnormal sound detection on the active noise reduction system according to the abnormal sound detection method described in any one of the first aspects;
[0053] In response to the presence of a stable single-frequency abnormal sound in the active noise reduction system, the active noise reduction system is restarted.
[0054] In a third aspect, an abnormal sound detection device is provided, which is applied to an active noise reduction system, wherein the active noise reduction system includes a microphone and a speaker; the abnormal sound detection device is used to implement the abnormal sound detection method described in any one of the first aspects; the abnormal sound detection method includes:
[0055] An acquisition module, used to acquire a first noise signal during the operation of the active noise reduction system; the first noise signal is a noise signal of a noise source collected by the microphone and / or a control signal of the speaker;
[0056] A detection module is used to perform Fourier transform on the first noise signal with a target frequency resolution, and perform abnormal sound detection on the active noise reduction system according to the Fourier transform result; wherein the target frequency resolution is obtained by optimization, and the optimization goal is to accurately identify the abnormal sound and minimize the calculation amount of the abnormal sound detection.
[0057] In a fourth aspect, an active noise reduction system is provided, comprising a memory, a processor, and a computer program stored in the memory and used to run on the processor, wherein when the processor executes the computer program, the method described in any one of the first aspect or the second aspect is implemented.
[0058] On the basis of being in accordance with the common sense in the art, the above-mentioned preferred conditions can be arbitrarily combined to obtain the preferred embodiments of the present disclosure.
[0059] The positive progressive effect of the present disclosure lies in that: the present disclosure performs spectrum analysis on the first noise signal according to the target frequency resolution obtained by optimization, which, on the one hand, can provide a data basis that conforms to the characteristics of the abnormal sound, eliminate interference, and thus improve the accuracy of abnormal sound detection; on the other hand, the spectrum analysis of the first noise signal is performed at the target frequency resolution, and the calculation amount of the abnormal sound detection is minimized, thereby being able to improve the abnormal sound detection rate and meet the timeliness requirements. BRIEF DESCRIPTION OF THE DRAWINGS
[0060] Figure 1 A flowchart of a method for detecting abnormal sound provided by an exemplary embodiment of the present disclosure;
[0061] Figure 2a A schematic diagram of a time domain curve of a stable single-frequency abnormal sound, howling, and normal noise provided by an exemplary embodiment of the present disclosure;
[0062] Figure 2b A schematic diagram of a frequency domain curve of a stable single-frequency abnormal sound, howling, and normal noise provided by an exemplary embodiment of the present disclosure;
[0063] Figure 3 A schematic diagram of spectrum lines corresponding to different frequency resolutions provided by an exemplary embodiment of the present disclosure;
[0064] Figure 4 A flowchart of optimizing a preset power range in an abnormal sound detection method provided by an exemplary embodiment of the present disclosure;
[0065] Figure 5 A schematic diagram comparing the curve effects of an abnormal sound detection method with and without A-weighting processing provided by an exemplary embodiment of the present disclosure;
[0066] Figure 6 A flowchart of another abnormal sound detection method provided by an exemplary embodiment of the present disclosure;
[0067] Figure 7 A flowchart of another abnormal sound detection method provided by an exemplary embodiment of the present disclosure;
[0068] Figure 8 A schematic diagram of a module of an abnormal sound detection device provided by an exemplary embodiment of the present disclosure;
[0069] Fig. 9 A schematic diagram of a module of an active noise reduction system provided by an exemplary embodiment of the present disclosure. DETAILED DESCRIPTION
[0070] The present disclosure is further described below by way of examples, but the present disclosure is not limited to the scope of the examples.
[0071] Prefixes such as "first" and "second" are used in the embodiments of the present disclosure only to distinguish different description objects, and have no limiting effect on the position, order, priority, quantity or content of the described objects. The use of prefixes such as ordinal numbers to distinguish description objects in the embodiments of the present disclosure does not constitute a limitation on the described objects. For the statement of the described objects, please refer to the description in the context of the claims or embodiments, and no unnecessary limitation should be constituted due to the use of such prefixes. In addition, in the description of the present embodiment, unless otherwise specified, the meaning of "plurality" is two or more.
[0072] Figure 1 A flowchart of an unusual sound detection method provided by an exemplary embodiment of the present disclosure is provided. The unusual sound detection method is applied to an active noise reduction system, and the active noise reduction system includes components such as a microphone, a speaker, and a controller.
[0073] The abnormal sound detected in the embodiment of the present disclosure refers to an abnormal sound emitted by the speaker of the active noise reduction system due to some faults or interference when the active noise reduction system is running / working, which affects the user experience. The abnormal sound includes howling and stable single-frequency abnormal sound (or stable single-frequency sound).
[0074] The characteristics of howling caused by active noise reduction are: Figure 2a and Figure 2b There is a prominent peak in the noise spectrum. The amplitude of the highest point of the peak is significantly higher than the adjacent frequency, and the peak is generally much larger than the peak of the stable single-frequency abnormal sound, accounting for a large proportion of the overall noise power. From the beginning of the howling, its power amplitude in the time domain gradually increases from small to large. The ratio of the maximum power to the minimum power is generally more than 5 times, and the amplitude is generally close to or even exceeds the maximum sound power of the speaker itself.
[0075] Different from howling, the characteristics of stable single-frequency noise / stable single-frequency sound are: there is a prominent peak in the noise spectrum, the amplitude of the highest point of the peak (noise peak) is higher than the amplitude corresponding to the adjacent frequency, and the power of the peak accounts for a small proportion of the total noise power, that is, the power of the peak accounts for a proportion of the total noise power that is less than or equal to the power threshold, where the power threshold is calibrated according to experimental data. Figure 2a As shown by the blue curve, the peak bandwidth does not exceed 50Hz. Figure 2b When a stable single-frequency abnormal sound is generated, its power amplitude in the time domain is relatively stable, the power changes little, the ratio of maximum power to minimum power will not exceed 3 times, and the amplitude is generally much smaller than the maximum sound power of the speaker itself.
[0076] The disclosed embodiment mainly detects whether the active noise reduction system has a stable single-frequency abnormal sound. Figure 1 , the abnormal sound detection method comprises the following steps:
[0077] Step 101: Acquire a first noise signal during the operation of an active noise reduction system.
[0078] The first noise signal can be a noise signal emitted by a noise source collected by the first microphone. The noise signal emitted by the noise source is subsequently used as the basic data for abnormal sound detection. The first microphone is deployed near the noise source. It should be noted that the first noise signal can be the collection result of one first microphone, or it can be a combination of the collection results of multiple first microphones. At this time, the first noise signal collected by the first microphone is the superposition result of the original noise signal emitted by the noise source and the feedback signal.
[0079] The active noise reduction system can be used to reduce the noise of appliances such as range hoods and washing machines. Taking the noise reduction of a range hood as an example, the noise source can be, for example, a fan of the range hood, and the first microphone for collecting noise signals is deployed near the fan.
[0080] The first noise signal can be a control signal of a speaker, which is then used as basic data for abnormal sound detection. It should be noted that the control signal of the speaker can be obtained directly by obtaining the output result of the controller, or it can be collected by a second microphone deployed near the speaker included in the active noise reduction system. The first noise signal can be a control signal of one speaker, or a combination of control signals of multiple speakers.
[0081] The first noise signal may also be a combination of a noise signal of a noise source and a control signal of a speaker. The specific combination is not particularly limited in the embodiment of the present disclosure.
[0082] Step 102: Perform Fourier transform on the first noise signal at a target frequency resolution, and perform abnormal sound detection on the active noise reduction system according to the Fourier transform result.
[0083] The target frequency resolution is obtained by optimization, and the optimization goal is to accurately identify the abnormal sound and minimize the calculation amount of the abnormal sound detection.
[0084] The so-called accurate recognition of abnormal sounds means that the recognition accuracy is greater than the accuracy threshold and / or the false recognition rate is less than or equal to the false recognition threshold. Among them, the accuracy threshold and the false recognition threshold can be set according to actual conditions.
[0085] In one embodiment, in order to increase the algorithm rate, a fast Fourier transform (FFT) is used to perform Fourier transform on the first noise signal.
[0086] The frequency resolution determines the amount of calculation in the process of abnormal sound detection. When the frequency resolution is 1Hz, it means that Fourier transform is performed every 1s. Assuming that the amount of calculation each time is K1, when the frequency resolution is increased to 4Hz, it means that Fourier transform is performed every 1 / 4s. At this time, the amount of calculation each time is K2, the amount of data within 1 / 4s, K1=4•K2. The frequency resolution of 4Hz reduces the amount of data by 4 times compared with the frequency resolution of 1Hz, which greatly speeds up the calculation speed of abnormal sound detection. The following table shows the relevant parameters of Fourier transform under different frequency resolutions when the test time is 4s and the frequency is 0-100Hz. It can be seen that the total number of calculations within 4s within 0-100Hz is the same, but the amount of calculation of each Fourier transform is different. It can be concluded from the table that as the frequency resolution increases, the amount of data for each Fourier transform decreases.
[0087]
[0088] In addition, as the frequency resolution increases, the amplitude of the amplitude-frequency diagram after Fourier transform will increase. That is to say, when the frequency resolution increases, the number of spectral lines decreases, and the power of each spectral line is the sum of the powers of the spectral lines before the frequency resolution increase.
[0089] In this embodiment, the spectrum analysis of the first noise signal is performed according to the target frequency resolution obtained by optimization. On the one hand, it can provide a data basis that conforms to the characteristics of the abnormal sound, eliminate interference, and thus improve the accuracy of abnormal sound detection; on the other hand, the spectrum analysis of the first noise signal is performed at the target frequency resolution, and the calculation amount of the abnormal sound detection is minimized, thereby being able to improve the abnormal sound detection rate and meet the timeliness requirements.
[0090] In one embodiment, the step of determining the target frequency resolution includes: obtaining abnormal samples, iteratively updating the frequency resolution, and determining the frequency resolution corresponding to accurate identification of abnormal sounds and the minimum calculation amount of abnormal sound detection as the target frequency resolution.
[0091] The abnormal sample is a noise signal containing an abnormal sound. The abnormal sample can be collected through experiments, or it can be the result of superimposing a normal sample (a noise signal without an abnormal sound) and an abnormal sound.
[0092] In one embodiment, the optimization algorithm may include, but is not limited to, at least one of the following algorithms: a genetic algorithm, a simulated annealing algorithm, a particle swarm algorithm, a Bayesian optimization algorithm, and a hill climbing algorithm.
[0093] The following example further illustrates the process of determining the target frequency resolution:
[0094] Based on the abnormal sound samples, the frequency band where the abnormal sound occurs is known, such as the existence of stable single-frequency abnormal sounds at frequencies fa, fb, and fc in the abnormal sound samples.
[0095] S1. Assuming that the initial frequency resolution is 1 Hz, after increasing the frequency resolution to 32 Hz, perform FFT on the abnormal sound samples and perform abnormal sound detection based on FFT.
[0096] S2. If the frequency resolution of 32Hz can accurately identify the presence of stable single-frequency abnormal sounds in the abnormal sound samples at fa, fb, and fc, it means that increasing the frequency resolution to 32Hz is sufficient to determine the abnormal sounds. If the frequency resolution of 32Hz cannot accurately identify the abnormal sounds at all frequencies fa, fb, and fc in the abnormal sound samples and / or there are misjudgments (for example, other frequencies are identified as abnormal sounds), it means that although increasing the frequency resolution to 32Hz reduces the amount of calculation, it is not enough to determine all the abnormal sound points, and the frequency resolution needs to be lowered.
[0097] S3. If the frequency resolution is sufficient to identify the abnormal sound, select 32 Hz as the frequency resolution and determine the corresponding calculation amount. At this time, the calculation amount has been reduced to 1 / 32 compared with the initial calculation amount when the frequency resolution is 1 Hz. If the frequency resolution is not sufficient to determine all the active noise reduction abnormal sound points, continue to repeat step 2 until the frequency resolution that accurately identifies the abnormal sound and minimizes the calculation amount of the abnormal sound detection is found.
[0098] In one embodiment, the first noise signal meets the sampling frequency requirements, total sampling time requirements, etc. of the Fourier transform processing. For example, the sampling frequency f s ≥2f U , where f U The maximum analysis frequency of the active noise reduction system during the noise signal acquisition process (for example, 12800Hz), the maximum analysis frequency is greater than or equal to the maximum frequency (abnormal sound frequency) of the abnormal sound that may occur f ANC For example, the frequencies of abnormal sound may be 700Hz, 800Hz and 900Hz. ANC is 900Hz, f U ≥900Hz. Total sampling time T≥ single Fourier transform time T FFT . T FFT It is determined by the frequency resolution Δf required for abnormal sound detection, Δf≤100 Hz. Preferably, Δf≤20 Hz.
[0099] Among them, the maximum analysis frequency of active noise reduction refers to the target control frequency of active noise reduction. Active noise reduction only analyzes noise in a certain frequency range and does not control noise that is not in the corresponding frequency range. The maximum analysis frequency of active noise reduction control ranges from [600Hz, 2000Hz]. Preferably, the maximum analysis frequency of active noise reduction control is [600Hz, 1000Hz].
[0100] For example, in a scenario where the active noise reduction system is used to reduce the noise of a range hood, the noise of the range hood is distributed between 0 and 20,000 Hz, while the frequency range of the analysis frequency of the active noise reduction may be only 0 to 1,000 Hz. Correspondingly, the abnormal sound caused by the active noise reduction system can only be between 0 and 1,000 Hz, so it is sufficient to set the maximum analysis frequency of the active noise reduction to 1,000 Hz.
[0101] In one embodiment, the step of obtaining the first noise signal during the operation of the active noise reduction system includes: obtaining the first noise signal during the operation of the active noise reduction system at a target sampling frequency. The target sampling frequency is determined according to the abnormal sound frequency and / or the computing power of the analysis device of the first noise signal; the abnormal sound frequency is determined according to the noise sample containing the abnormal sound. The noise sample can be obtained through experiments, or it can be obtained by superimposing abnormal sounds on noise samples without abnormal sounds. The analysis device of the first noise signal is also a device that analyzes the first noise signal to realize abnormal sound detection, for example, it can be an active noise reduction chip / controller.
[0102] Here is a way to determine the target sampling frequency:
[0103] Method for determining the minimum value of the target sampling frequency: By analyzing the noise samples containing abnormal sounds, it can be seen that the maximum frequency of abnormal sounds (abnormal sound frequency) is generally f ANC According to the Nyquist sampling theorem, in order to avoid aliasing, the target sampling frequency should be at least twice the frequency of the aliasing tone. Therefore, the minimum value of the target sampling frequency is f s_min =2f ANC .
[0104] Method for determining the maximum value of the sampling frequency: Analyzing the first noise signal includes performing Fourier transform on the first noise signal. The computing power of the analysis device can be characterized by the number of data points of the first noise signal each time the Fourier transform is performed. The maximum value of the target sampling frequency can be obtained by inferring the maximum computing power. Assume that the number of data points allowed to be transformed by the analysis device in one Fourier transform is N, the time of one Fourier transform is 1 / Δf, and the sampling frequency f s Indicates sampling per second s data points, so the number of data points allowed to be transformed by a Fourier transform is f s 1 / Δf, that is, N=f s 1 / Δf, assuming that the maximum computing power of the analysis device is N, and the maximum number of data points per Fourier transformation is N max , we can get f s_max =N max ·Δf.
[0105] Therefore, the target sampling frequency f of the first noise signal is s_res The value range is [f s_min, f s_max ]. In the related art, in order to improve the control accuracy of active noise reduction, the original signal collected by the active noise reduction controller has a higher sampling frequency f s , usually f s In this embodiment, the target sampling frequency f is less than 12000 Hz. s_res A first noise signal is obtained.
[0106] In this embodiment, the first noise signal is obtained at a target sampling frequency determined according to the computing power of the analysis device of the abnormal sound frequency and / or the first noise signal, and is used as the data basis for abnormal sound detection. This can eliminate interference and provide a data basis that meets the characteristics of the abnormal sound, thereby improving the accuracy of abnormal sound detection and further reducing the amount of data analysis, thereby increasing the abnormal sound detection rate and meeting timeliness requirements.
[0107] In one embodiment, in step 101 , an original noise signal is first acquired, and the original noise signal is resampled at a target sampling frequency to obtain a first noise signal.
[0108] In one embodiment, step 102 includes: determining the total spectral power amplitude of the first noise signal (the total power of the first noise signal) according to the Fourier transform result; in response to the total spectral power amplitude being within a preset power range, determining that the active noise reduction system has a stable single-frequency abnormal sound; wherein the lower limit value of the preset power range is greater than the maximum power of the noise signal of the noise source, and the upper limit value of the preset power range is less than the power of the first noise signal when howling occurs in the active noise reduction system.
[0109] Perform Fourier transform on the first noise signal to obtain the spectrum data of the first noise signal (Fourier transform result). Through the analysis of the spectrum data, it is found that when howling and stable single-frequency abnormal sound occur, there is a signal peak value on the spectrum that is significantly higher than the surrounding frequency, such as Figure 3 The peak value near 95Hz of the red dotted line and the peak value near 109Hz of the blue thin solid line are shown. Through filtering and playback, it is found that the noise corresponding to the peak frequency is the sound that produces howling and stable single-frequency abnormal sound. The normal signal ( Figure 3 There is no obvious abrupt noise peak in the middle green thick solid line, so the spectrum analysis of the first noise signal can accurately identify the abnormal sound. Figure 3 The AutoPower Mono in indicates the curve name.
[0110] It should be noted that, when the first noise signal is the noise signal of the noise source, the upper limit value is less than the power of the noise signal collected by the first microphone when the active noise reduction system howls; when the first noise signal is the control signal of the speaker, the upper limit value is less than the power of the control signal of the speaker when the active noise reduction system howls; when the first noise signal is a combination of the noise signal of the noise source and the control signal of the speaker, the upper limit value is less than the power combination of the two when the active noise reduction system howls; the upper limit value can also be approximately the sound value corresponding to the maximum radiation power of the speaker.
[0111] In this embodiment, by performing spectrum analysis on the first noise signal, the total spectrum power amplitude related to the abnormal sound in the first noise signal is determined, and then the abnormal sound can be accurately identified according to the total spectrum power amplitude, thereby improving the abnormal sound detection accuracy of the active noise reduction system.
[0112] The preset power range can be an empirical value, such as [74.5dB, 91dB]. When the active noise reduction system howls, the sound emitted by the speaker reaches its maximum radiation power, and the sound is usually greater than 91dB (A). The upper limit of the preset power range is set to 91dB. Taking the range hood equipped with the active noise reduction system as an example, the maximum noise (sound pressure level) of the range hood when the active noise reduction system is not started is E AM It is about 70dB(A). So generally the preset power range is E AT The lower limit is set to a value greater than 70dB(A), for example, E AT The lower limit value is 74.5 dB(A). When the total amplitude of the spectrum power of the first noise signal is detected to exceed 74.5 dB(A), it means that the active noise reduction not only fails to reduce noise, but amplifies the sound. The corresponding reason is that abnormalities such as howling or stable single-frequency abnormal sound are generated. However, the total amplitude of the spectrum power does not exceed 91dB, which means that the active noise reduction system has a stable single-frequency abnormal sound.
[0113] The preset power range can also be obtained by dynamic optimization to improve the accuracy of abnormal sound detection. The following introduces a method that can be implemented by dynamically optimizing the preset power range. Figure 4 , the preset power range is determined by:
[0114] Step 100-11: Obtain abnormal signal samples and normal samples.
[0115] The abnormal sample is a noise signal containing an abnormal sound, and the normal sample is a noise signal without an abnormal sound. The abnormal sample can be collected through experiments, or it can be the result of superimposing a normal sample with a noise signal whose frequency and amplitude match the stable single-frequency abnormal sound.
[0116] The number of abnormal samples and normal samples can be set according to actual needs. It can be understood that the more samples there are, the better the optimization can be performed.
[0117] Step 100-12: Construct the objective function.
[0118] The objective function is constructed based on at least one of the following parameters: the number of missed abnormal samples, the number of falsely detected normal samples, the percentage of missed abnormal samples, and the percentage of falsely detected normal samples.
[0119] The design principle of the objective function is to find a combination of judgment parameters and values so that the preset power range obtained by optimization can detect all abnormal sounds, while normal signals are not mistakenly judged as abnormal sounds, and at the same time have good robustness. The objective function of parameter optimization can be set as: (1) (number of missed abnormal samples + number of false normal samples) / total number of samples. In this case, the smaller the objective function, the better; (2) (abnormal sound judgment strength of normal samples) / (abnormal sound judgment strength of abnormal signals). In this case, the smaller the objective function, the better.
[0120] Among them, the abnormal sound judgment strength = abnormal sample / normal sample conforms to the degree of abnormal sound judgment. If the total abnormal sound threshold is 100, and a value above 100 is an abnormal sound, then if the value of the normal sample is 30, the abnormal sound strength is 0.3, which is the abnormal sound judgment strength of the normal sample. The smaller the value, the farther the normal sound is from the standard of the abnormal sound, the greater the probability of the normal sound, and the higher the accuracy of the detection algorithm. If the value of the abnormal sound sample is 200, the abnormal sound strength is 2, which is the abnormal sound judgment strength of the abnormal signal. The larger the value, the more likely the sound is an abnormal sound, the greater the probability of it being an abnormal sound, and the higher the accuracy of the detection algorithm. Therefore, the smaller the value of (normal sample abnormal sound judgment strength) / (abnormal signal abnormal sound judgment strength), the higher the accuracy of the detection algorithm, and the better the corresponding parameter combination for detection.
[0121] The strength of abnormal sound determination can also be defined in other ways, for example: the strength of abnormal sound determination of normal samples = the number of normal samples mistakenly determined as abnormal sounds / the total number of normal samples; the strength of abnormal sound determination of abnormal signals = the number of abnormal samples accurately determined as abnormal sounds / the total number of abnormal samples.
[0122] In addition to the above objective function, other similar calculation formulas that conform to the objective function design principles may also be used, and the embodiments of the present disclosure are not particularly limited thereto.
[0123] Step 100-13: Optimize the objective function within a preset optimization range of the upper limit value and / or the lower limit value to determine a final preset power range.
[0124] It should be noted that the upper limit value and the lower limit value of the preset power range may be optimized simultaneously or separately, and this is not particularly limited in the embodiments of the present disclosure.
[0125] In one embodiment, the preset optimization range is [74dB, 91dB], that is, the preset power range searches for the optimal solution in [74dB, 91dB] to minimize the objective function; or, the preset optimization range of the lower limit value is [70dB, 80dB], that is, the lower limit value searches for the optimal solution in [70dB, 80dB]; the preset optimization range of the upper limit value is [85dB, 95dB], that is, the upper limit value searches for the optimal solution in [85dB, 95dB].
[0126] In this embodiment, an optimization range of a preset power range is provided, so that the optimization meets the physical characteristics of certain abnormal sounds, and the optimization accuracy and optimization efficiency can be improved.
[0127] In one embodiment, the optimization algorithm includes at least one of the following algorithms: genetic algorithm, simulated annealing algorithm, particle swarm algorithm, Bayesian optimization algorithm, and hill climbing algorithm.
[0128] In one embodiment, the abnormal sound judgment condition includes: in response to the difference between the total amplitude of the spectrum power and the noise peak being less than the first difference threshold, determining that the active noise reduction system has a stable single-frequency abnormal sound. Step 102 includes: performing Fourier transform on the first noise signal, and determining the total amplitude of the spectrum power and the noise peak of the first noise signal according to the Fourier transform result; in response to the difference between the total amplitude of the spectrum power and the noise peak being less than the first difference threshold, determining that the active noise reduction system has a stable single-frequency abnormal sound.
[0129] The first difference threshold ΔE PT It can be an empirical value, such as 4 dB. When the active noise reduction system is working, if the total amplitude of the spectrum power of the first noise signal E is measured A The first noise signal is composed of the sum of all sound sources at frequencies between 0 and 6000 Hz. Through spectrum analysis, the maximum peak (noise peak) in the spectrum appears at 500 Hz, and its noise peak E P is 62dB(A). A and E P Convert dB to success rate respectively, 65dB: power 1=10 (0.1*65) = 3162277; 62dB: Power 2 = 10 (0.1*62)= 1584893; Power 2 / Power 1≈0.5 = 50%. At this time, the sound at a frequency of 500Hz accounts for 50% of the total amplitude of the spectrum power, and the power of other sound frequencies combined only accounts for 50%. Therefore, in the spectrum of this sound, the sound at 500Hz is obviously louder than the sounds of other frequencies. At this time, the active noise reduction system will emit a very unpleasant single-frequency sound, that is, an abnormal sound. If the sound is measured in dB, then E A -E P =65-62=3dB<ΔE PT , determine the existence of a stable single-frequency abnormal sound.
[0130] The first difference threshold value may also be obtained based on dynamic optimization to improve the accuracy of abnormal sound detection. The specific optimization process of the first difference threshold value is similar to the optimization process of the preset power range, which will not be repeated here.
[0131] In one embodiment, the second preset optimization range of the first difference threshold is [1 dB, 10 dB].
[0132] In one embodiment, the abnormal sound detection condition includes determining that there is a stable single-frequency abnormal sound in the active noise reduction system in response to the presence of a noise segment in the first noise signal whose power ratio is greater than or equal to a ratio threshold. The power ratio is the ratio of the power of the noise segment to the total amplitude of the spectrum power.
[0133] The noise segment is the result of segmenting the noise signal. The segmentation algorithm can be, but is not limited to, one of the following: 1 / n octave method, Bark bandwidth method, equal bandwidth method, etc. The first two are preferred, where n in 1 / n octave is a non-zero natural number such as 1, 2, 3, etc., taking into account both detection accuracy and computing power, preferably n=3. The segmentation algorithm can also use the segmentation algorithm provided in the following embodiments.
[0134] The percentage threshold may be an empirical value, or may be obtained by dynamic optimization to improve the accuracy of abnormal sound detection. The specific optimization process of the percentage threshold is similar to the optimization process of the preset power range, and will not be described here.
[0135] In one embodiment, the abnormal sound detection condition includes determining that there is a stable single-frequency abnormal sound in the active noise reduction system in response to the difference between the total amplitude of the spectrum power and the filter power being greater than or equal to a second difference threshold. The filter power is the result of filtering out the noise peak from the total amplitude of the spectrum power.
[0136] The second difference threshold may be an empirical value, or may be obtained by dynamic optimization to improve the accuracy of abnormal sound detection. The specific optimization process of the second difference threshold is similar to the optimization process of the preset power range, and will not be described here.
[0137] In one embodiment, the alien sound detection condition includes determining that a stable single-frequency alien sound exists in response to the presence of a signal segment having a power difference with an adjacent signal segment greater than a third difference threshold.
[0138] The third difference threshold may be an empirical value, or may be obtained by dynamic optimization to improve the accuracy of abnormal sound detection. The specific optimization process of the third difference threshold is similar to the optimization process of the preset power range, and will not be described here.
[0139] The above abnormal sound detection conditions can be used in combination. As long as one of the other conditions is met, it is determined that the active noise reduction system has a stable single-frequency abnormal sound; otherwise, it is determined that the active noise reduction system does not have a stable single-frequency abnormal sound.
[0140] In one embodiment, the step of determining the total spectral power amplitude of the first noise signal according to the Fourier transform result includes: performing A-weighting processing on the Fourier transform result, and determining the total spectral power amplitude of the first noise signal according to the A-weighting processing result.
[0141] At present, the definition of abnormal sound is mainly based on people's subjective feelings. Therefore, although some frequencies are loud, they can still be ignored if the human ear is not sensitive to sounds in this frequency band (such as low-frequency sounds). In the abnormal sound detection process, the Fourier transform result of the first noise signal is first A-weighted based on the sensitivity of the human ear, and then the A-weighted processing result is used for abnormal sound detection. Since A-weighting can better reflect the human ear's auditory perception, this detection method is easier to detect the real abnormal sound, with better detection effect, and more in line with the user's perception and judgment of abnormal sound.
[0142] See also Figure 5 The green solid line is the spectrum of the first noise signal without A-weighting, in which there are many low-frequency components below 100 Hz that are insensitive to human ears. The sound signal of the first noise signal after A-weighting is as follows Figure 5 As shown by the red dotted line in the middle, the proportion of the total power of the low-frequency components that the human ear is not sensitive to is reduced. A -E P The value of E is changed from 3.2dB to 2.5dB, which is more in line with the judgment of abnormal sound. A -E P Under the condition of <5dB, the detection is more accurate and more in line with the perception of the human ear.
[0143] It should be noted that, since it is easier to determine abnormal sounds after A-weighting processing, the threshold is adjusted according to suitability after A-weighting processing. For example, the first difference threshold can be smaller than the first difference threshold without A-weighting processing, for example, the first difference threshold without A-weighting processing is 5dB, and the first difference threshold after A-weighting processing is 4dB.
[0144] In one embodiment, the abnormal sound is detected by multiple judgments, and step 103 includes: in response to the number of occurrences of the total amplitude of the spectrum power within the preset power range being greater than the number threshold, determining that the active noise reduction system has a stable single-frequency abnormal sound.
[0145] When performing abnormal sound detection on the spectrum data of the first noise signal, if the collected first noise signal contains a noise interference signal, or the first noise signal used for detection itself is not a completely steady-state signal, there is a certain fluctuation characteristic over time. These reasons lead to differences in the spectrum data obtained at different times, which in turn causes misjudgment or wrong judgment of abnormal sound detection.
[0146] In order to reduce or avoid misjudgment and wrong judgment, in this embodiment, whenever the total amplitude of the spectrum power is within the preset power range, it is determined that the active noise reduction system has an abnormality, and the number of abnormal occurrences is counted by 1. Based on the noise signal collected later, it is determined that the active noise reduction system has an abnormality again, and the number of abnormal occurrences is counted by 1 again. When the number of occurrences is greater than the number threshold, it is determined that the active noise reduction system has a stable single-frequency abnormal sound. Misjudgment and wrong judgment can be effectively reduced or avoided through multiple judgments.
[0147] The number threshold can be set according to actual needs. It can be an empirical value, such as 3 times; it can also be dynamically determined through target optimization to improve the accuracy of abnormal sound detection. The target optimization process of the number threshold is similar to the preset power range, which will not be repeated here.
[0148] Many abnormal sound frequencies have certain fluctuations, and the amplitude of the corresponding frequency has a great influence. When using the spectrum to identify abnormal sounds, there are misjudgments and missed judgments. Moreover, people's perception and judgment of abnormal sounds are based on the overall perception of the power in a frequency band, rather than the power at a single frequency. Therefore, integrating the noise spectrum according to a certain rule and using the power of the corresponding frequency band after integration to judge the abnormal sound is more in line with people's actual feelings.
[0149] Based on this, in one embodiment, the Fourier transform result (spectrum) of the first noise signal is segmented to obtain at least two noise segments, and the total spectrum power amplitude is determined according to the integration of each noise segment. The total spectrum power amplitude here is the total spectrum power amplitude of the first noise signal globally.
[0150] The segmentation algorithm may be, but is not limited to, one of the following: 1 / n octave method, Bark bandwidth method, equal bandwidth method, etc. The first two are preferred, where n in 1 / n octave is a non-zero natural number such as 1, 2, 3, etc., taking into account both detection accuracy and computing power, and n=3 is preferred.
[0151] In one embodiment, each noise segment is subjected to abnormal sound detection. Specifically, the Fourier transform result (spectrum) of the first noise signal is processed in segments, the total spectrum power amplitude of each noise segment is calculated, and the abnormal sound detection is performed on each noise segment. The total spectrum power amplitude is the local spectrum power amplitude of each noise segment.
[0152] In one embodiment, before the Fourier transform result of the first noise signal is segmented, the Fourier transform result is A-weighted, and then the A-weighted result is segmented.
[0153] Equipment equipped with an active noise reduction system, such as a range hood, may also produce a sound with a spectrum similar to the stable single-frequency sound caused by the active noise reduction system, which we call "normal abnormal sound". Although this "normal abnormal sound" may also be offensive to users, it is different from the stable single-frequency sound caused by the active noise reduction system and is not a "by-product" of the active noise reduction system. It cannot be reduced by turning off or adjusting the active noise reduction system.
[0154] Based on this, in one embodiment, if the detection result of step 103 is that there is a stable single-frequency abnormal sound, in order to exclude the possibility that the abnormal sound is caused by a device equipped with an active noise reduction system rather than the active noise reduction system, the active noise reduction system is first turned off, and a second noise signal during the process of turning off the active noise reduction system is obtained. The second noise signal is Fourier transformed, and the total amplitude of the spectral power is calculated to determine whether there is still a stable single-frequency abnormal sound; if the judgment result is that there is a stable single-frequency abnormal sound in the second noise signal, and the abnormal sound frequency of the second noise signal is the same as or similar to that of the first noise signal, it is determined that the abnormal sound is caused by the device rather than the active noise reduction system; otherwise, it is determined that there is a stable single-frequency abnormal sound in the active noise reduction system.
[0155] Another implementation method for eliminating the interference of abnormal noise of the equipment on the abnormal noise detection of the active noise reduction system is provided below.
[0156] In one embodiment, see Figure 6 , before obtaining the first noise signal during the operation of the active noise reduction system, it also includes:
[0157] Step 100: When the active noise reduction system is not started, a second noise signal of a noise source is obtained, and the second noise signal is segmented so that the power proportion of each noise segment of the second noise signal is less than a proportion threshold, and a segmentation strategy of the second noise signal is determined.
[0158] The second noise signal is collected by the first microphone.
[0159] The number of noise segments obtained by segmenting the first noise signal is determined according to the power proportion of each noise segment. If the power proportion of each noise segment is not less than the proportion threshold, it is necessary to perform segmentation again. The frequency bands of each noise segment can be the same or different. The second noise signal is segmented so that the power proportion of each noise segment of the second noise signal is less than the proportion threshold, thereby eliminating the influence of the frequency band characteristics of the range hood itself on the judgment of abnormal sound.
[0160] The percentage threshold can be set according to the actual situation. The value range of the percentage threshold can be, for example, [40%, 50%]. Preferably, the percentage threshold is selected as 40%. 40% is selected to leave a margin to avoid being too close to 50% which may cause misjudgment. This is because: since the subsequent criterion for judging abnormal sounds is that if the percentage threshold exceeds 50%, it is judged that there is a stable single-frequency abnormal sound. Since it cannot be guaranteed that the power percentage of a certain noise segment is an accurate value, there may be a fluctuation of 5%. Therefore, if it is too close to 50%, it may cause the misjudgment of abnormal sounds due to fluctuation errors. When the percentage threshold is set to 40%, when the first power percentage is less than 40%, it means that the power of the noise segment has dropped to a low enough level to avoid misjudgment caused by power fluctuations in the noise segment.
[0161] The segmentation strategy includes the following parameters: the number of segments, the upper and lower frequency limits of each noise segment.
[0162] The following is the calculation method of power ratio, the formula is as follows:
[0163] H=1-e i / E;
[0164] ;
[0165] Among them, H represents the power ratio; e i represents the power corresponding to the ith noise segment in the noise signal; E represents the sum of the powers of the noise segments contained in the noise signal; X j (t) represents the amplitude of the jth sampling point of the noise signal in the time domain; N represents the number of sampling point noise segments contained in the noise signal. i The calculation method of is similar to E.
[0166] In step 102, the first noise signal is processed in segments according to the segmentation strategy, and Fourier transform is performed on the noise segments obtained by segmentation, or the Fourier transform result is processed in segments according to the segmentation strategy. Subsequently, the total spectral power amplitude of the first noise signal is determined globally according to the integral of each noise segment, or the total spectral power amplitude of each noise segment is determined separately, that is, the total local spectral power amplitude.
[0167] In this embodiment, the abnormal sound of the active noise reduction system can be quickly monitored, the influence of the frequency band characteristics of the range hood itself on the abnormal sound judgment can be eliminated, and accurate monitoring of the active noise reduction abnormal sound of the range hood can be achieved.
[0168] Another implementation method for eliminating the interference of abnormal noise of the equipment on the abnormal noise detection of the active noise reduction system is provided below.
[0169] In one embodiment, see Figure 7 , before obtaining the first noise signal during the operation of the active noise reduction system, it also includes:
[0170] Step 100-21: When the active noise reduction system is not started, obtain a second noise signal from a noise source.
[0171] Step 100-22: perform octave calculation on the second noise signal to determine a segmentation strategy that matches the second noise signal.
[0172] The n in the 1 / n octave is a non-zero natural number such as 1, 2, or 3. Taking both detection accuracy and computing power into consideration, n=3 is preferably selected.
[0173] Step 100-23: segment the second noise signal according to the segmentation strategy, and calculate the characteristic value of each noise segment of the second noise signal.
[0174] The characteristic value includes power and / or decibel value.
[0175] Step 100-24: determine a target gain of a first target noise segment whose eigenvalue is greater than an eigenvalue threshold.
[0176] After the first target noise segment is filtered by a bandpass filter with a target gain, the characteristic value of the first target noise segment is less than or equal to the characteristic value threshold.
[0177] The eigenvalue of the first target noise segment is greater than the eigenvalue threshold, indicating that the first target noise segment contains abnormal sound. The active noise reduction system was not started when the first target noise segment was collected, which means that the abnormal sound contained in the first target noise segment is caused by the device equipped with the active noise reduction system, rather than the active noise reduction system. In order to eliminate the interference of abnormal sound of the device during abnormal sound detection, it is necessary to filter the second target noise segment contained in the first noise signal to filter out the abnormal sound caused by the device. Among them, the frequency band of the second target noise segment includes the frequency band of the first target noise segment.
[0178] Taking the characteristic value as decibel value as an example, assuming that the decibel value of the first target noise segment before filtering is P, and the decibel value after filtering with the target gain bandpass filter is P' (P' is less than or equal to the characteristic value threshold), where P' is set to be 1dB greater than the decibel value of the adjacent frequency band, according to the formula: P'=20lg(E' / E0), the power of the first target noise segment after filtering can be calculated as E'. Based on the known power E before filtering and power E' after filtering of the first target noise segment, the target gain / pass rate of the first target noise segment is calculated to be α=E' / E.
[0179] It should be noted that the reason why P' is set to be 1dB greater than the decibel value of the adjacent frequency band instead of being equal is to make the abnormal sound monitoring more sensitive. Compared with the two, the former is more sensitive to abnormal sound judgment. When the judgment condition is 3dB higher than the adjacent frequency band, because it is 1 db higher than the adjacent frequency band, a smaller power is required to judge the stable single-frequency abnormal sound compared to the same as the adjacent frequency band. At this time, the sensitivity of stable single-frequency abnormal sound judgment is higher and the accuracy can be guaranteed at the same time.
[0180] In step 102, the first noise signal is segmented to obtain at least two noise segments, and a bandpass filter with a target gain is used to filter the second target noise segment contained in the first noise signal, and the unfiltered noise segment and the filtered noise segment in the first noise signal are respectively Fourier transformed.
[0181] The unfiltered noise segments in the first noise signal are other noise segments except the filtered noise segment (the second target noise segment).
[0182] The second target noise segment is filtered by a bandpass filter with a target gain, that is, α is multiplied by the amplitude of the second target noise segment to filter the abnormal sound caused by the device in the second target noise segment. The amplitude of other noise segments in the second noise signal except the second target noise segment is not attenuated, and the power pass rate of other noise segments is 100%.
[0183] In this embodiment, a bandpass filter with a target gain is used to filter the noise segment of the target noise segment in the first noise signal, so as to eliminate the interference of abnormal noise of the equipment on the abnormal noise detection of the active noise reduction system, and further improve the accuracy of abnormal noise detection.
[0184] An embodiment of the present disclosure also provides a control method for an active noise reduction system, the method comprising: performing abnormal sound detection on the active noise reduction system according to the abnormal sound detection method provided in any of the above embodiments; in response to the presence of a stable single-frequency abnormal sound in the active noise reduction system, restarting the active noise reduction system.
[0185] According to experience, the abnormal noise of the active noise reduction system is caused by certain faults or interferences. Restarting the active noise reduction system can weaken or eliminate the faults or interferences, improve the effectiveness of the active noise reduction system, and thus improve the user experience.
[0186] Corresponding to the aforementioned abnormal sound detection method embodiment, the present disclosure also provides an embodiment of an abnormal sound detection device.
[0187] Figure 8 A schematic diagram of a module of an abnormal sound detection device provided by an exemplary embodiment of the present disclosure, the device comprising:
[0188] An acquisition module 81 is used to acquire a first noise signal during the operation of the active noise reduction system; the first noise signal is a noise signal of a noise source collected by a microphone and / or a control signal of the speaker;
[0189] The detection module 82 is used to perform Fourier transform on the first noise signal with a target frequency resolution, and perform abnormal sound detection on the active noise reduction system according to the Fourier transform result; wherein the target frequency resolution is obtained by optimization, and the optimization goal is to accurately identify the abnormal sound and minimize the calculation amount of the abnormal sound detection.
[0190] Optionally, the step of determining the target frequency resolution includes:
[0191] Acquire an abnormal sample; the abnormal sample is a noise signal containing an abnormal sound;
[0192] The frequency resolution is iteratively updated, and the frequency resolution corresponding to the time when the abnormal sound of the abnormal sample is detected and the calculation amount is minimized is determined as the target frequency resolution.
[0193] Optionally, obtaining a first noise signal during operation of the active noise reduction system includes:
[0194] A first noise signal during the operation of the active noise reduction system is obtained at a target sampling frequency; the target sampling frequency is determined according to an abnormal sound frequency and / or a computing power of an analysis device of the first noise signal; and the abnormal sound frequency is determined according to a noise sample containing an abnormal sound.
[0195] Optionally, the target sampling frequency is greater than or equal to twice the frequency of the abnormal sound;
[0196] And / or, the computing power is characterized by a maximum number of data points for each Fourier transform performed by the analysis device, and the target sampling frequency is less than or equal to the product of the maximum number of data points and the frequency of the Fourier transform.
[0197] Optionally, the acquisition module is specifically used for:
[0198] An original noise signal is acquired, and the original noise signal is resampled at the target sampling frequency to obtain the first noise signal.
[0199] Optionally, the detection module is specifically used for:
[0200] Determine the total amplitude of the spectral power of the first noise signal according to the Fourier transform result;
[0201] In response to the total amplitude of the spectrum power being within a preset power range, it is determined that the active noise reduction system has a stable single-frequency abnormal sound; wherein the lower limit of the preset power range is greater than the maximum power of the noise signal of the noise source, and the upper limit of the preset power range is less than the power of the first noise signal when howling occurs in the active noise reduction system;
[0202] and / or, in response to the difference between the total amplitude of the spectrum power and the noise peak being less than a first difference threshold, determining that a stable single-frequency abnormal sound exists in the active noise reduction system;
[0203] And / or, in response to the presence of a noise segment in the first noise signal whose power ratio is greater than or equal to a ratio threshold, determining that a stable single-frequency abnormal sound exists in the active noise reduction system; the power ratio is the ratio of the power of the noise segment to the total amplitude of the spectrum power;
[0204] And / or, in response to the difference between the total amplitude of the spectrum power and the filter power being greater than or equal to a second difference threshold, determining that the active noise reduction system has a stable single-frequency abnormal sound; the filter power is the result of filtering out the noise peak from the total amplitude of the spectrum power;
[0205] And / or, in response to the presence of a signal segment having a power difference with an adjacent signal segment greater than a third difference threshold, it is determined that a stable single-frequency alien sound exists.
[0206] Optionally, the device further includes a preset power range determination module, configured to:
[0207] Acquire abnormal samples and normal samples; the abnormal samples are noise signals containing abnormal sounds, and the normal samples are noise signals not containing abnormal sounds;
[0208] Constructing an objective function; the objective function is constructed according to at least one of the following parameters: the number of missed detections of abnormal samples, the number of false detections of normal samples, the percentage of missed detections of abnormal samples, and the percentage of false detections of normal samples;
[0209] The objective function is optimized within a preset optimization range of the upper limit value and / or the lower limit value to determine a final preset power range.
[0210] Optionally, the device further includes a difference threshold determination module, configured to:
[0211] Acquire abnormal samples and normal samples; the abnormal samples are noise signals containing abnormal sounds, and the normal samples are noise signals not containing abnormal sounds;
[0212] Constructing an objective function; the objective function is constructed according to at least one of the following parameters: the number of missed detections of abnormal samples, the number of false detections of normal samples, the percentage of missed detections of abnormal samples, and the percentage of false detections of normal samples;
[0213] The objective function is optimized within a second preset optimization range of the difference threshold to determine a final difference threshold.
[0214] Optionally, the first preset optimization range is [74dB, 91dB];
[0215] And / or, the second preset optimization range is [1dB, 10dB];
[0216] And / or, the optimization algorithm includes at least one of the following algorithms: genetic algorithm, simulated annealing algorithm, particle swarm algorithm, Bayesian optimization algorithm, hill climbing algorithm;
[0217] And / or, the objective function is the ratio of the number of anomalies to the total number of samples, and the number of anomalies is the sum of the number of missed detections of the abnormal samples and the number of false detections of normal samples.
[0218] Optionally, the determination module is specifically used to: perform A-weighting processing on the Fourier transform result, and determine the total amplitude of the spectrum power of the first noise signal according to the A-weighting processing result;
[0219] And / or, the determination module is specifically used to: divide the Fourier transform result into at least two noise segments, and determine the total amplitude of the spectrum power according to the integration of each noise segment;
[0220] And / or, the detection module is specifically used to: in response to the number of occurrences of the total amplitude of the spectrum power within a preset power range being greater than a number threshold, determine that the active noise reduction system has a stable single-frequency abnormal sound.
[0221] Optionally, the acquisition module is further configured to acquire a second noise signal from a noise source when the active noise reduction system is not started;
[0222] The device also includes: a segmentation module, configured to perform segmentation processing on the second noise signal so that the power proportion of each noise segment of the second noise signal is less than a proportion threshold, and determine a segmentation strategy for the second noise signal;
[0223] The determination module is specifically used to: perform segmentation processing on the first noise signal according to the segmentation strategy, and perform Fourier transform on the noise segments obtained by segmentation.
[0224] Optionally, the acquisition module is further configured to acquire a second noise signal from a noise source when the active noise reduction system is not started;
[0225] The device further includes: a calculation module, configured to perform octave calculation on the second noise signal to determine a segmentation strategy that matches the second noise signal;
[0226] A segmentation module, used for segmenting the second noise signal according to the segmentation strategy, and calculating the characteristic value of each noise segment of the second noise signal; the characteristic value includes power and / or decibel value;
[0227] A gain module, configured to determine a target gain of a first target noise segment whose characteristic value is greater than a characteristic value threshold; and the characteristic value of the first target noise segment after filtering the first target noise segment with a bandpass filter of the target gain is less than or equal to the characteristic value threshold;
[0228] Determine the module specifically for:
[0229] The first noise signal is segmented to obtain at least two noise segments, and a second target noise segment contained in the first noise signal is filtered using a bandpass filter with a target gain, and an unfiltered noise segment and a filtered noise segment in the first noise signal are respectively Fourier transformed; wherein a frequency band of the second target noise segment includes a frequency band of the first target noise segment.
[0230] For the device embodiment, since it basically corresponds to the method embodiment, the relevant parts can refer to the partial description of the method embodiment. The device embodiment described above is only illustrative, wherein the units described as separate components may or may not be physically separated, and the components as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the disclosed solution.
[0231] Fig. 9 This is a structural schematic diagram of an active noise reduction system shown in an example embodiment of the present disclosure. The active noise reduction system includes a memory, a processor, and a computer program stored in the memory and used to run on the processor. When the processor executes the computer program, the abnormal sound detection method or control method described in any of the above embodiments is implemented. Fig. 9 The active noise reduction system 90 shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.
[0232] like Fig. 9As shown, the components of the active noise reduction system 90 may include but are not limited to: the at least one processor 91 mentioned above, the at least one memory 92 mentioned above, and a bus 93 connecting different system components (including the memory 92 and the processor 91).
[0233] The bus 93 includes a data bus, an address bus, and a control bus.
[0234] The memory 92 may include a volatile memory, such as a random access memory (RAM) 921 and / or a cache memory 922 , and may further include a read only memory (ROM) 923 .
[0235] The memory 92 may also include a program tool 925 (or utility) having a set (at least one) of program modules 924, such program modules 924 including but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment.
[0236] The processor 91 executes various functional applications and data processing by running the computer program stored in the memory 92, such as the abnormal sound detection method or control method provided in any of the above embodiments.
[0237] The active noise reduction system 90 may also communicate with one or more external devices 94 (e.g., a range hood processor, etc.). Such communication may be performed via an input / output (I / O) interface 95. In addition, the active noise reduction system 90 may also communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) via a network adapter 96. As shown, the network adapter 96 communicates with other modules of the active noise reduction system 90 via a bus 93. It should be understood that, although not shown in the figure, other hardware and / or software modules may be used in conjunction with the active noise reduction system 90, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID (disk array) systems, tape drives, and data backup storage systems, etc.
[0238] It should be noted that although several units / modules or sub-units / modules of the active noise reduction system are mentioned in the above detailed description, this division is merely exemplary and not mandatory. In fact, according to the embodiments of the present disclosure, the features and functions of two or more units / modules described above can be embodied in one unit / module. Conversely, the features and functions of one unit / module described above can be further divided to be embodied by multiple units / modules.
[0239] The embodiments of the present disclosure further provide a computer-readable storage medium having a computer program stored thereon. When the program is executed by a processor, the abnormal sound detection method or control method provided in any of the above embodiments is implemented.
[0240] The readable storage medium may include but is not limited to: a portable disk, a hard disk, a random access memory, a read-only memory, an erasable programmable read-only memory, an optical storage device, a magnetic storage device or any suitable combination of the above.
[0241] The embodiment of the present disclosure further provides a computer program product, including a computer program, wherein when the computer program is executed by a processor, the computer program implements any of the above-mentioned abnormal sound detection methods or control methods.
[0242] Among them, the program code for executing the computer program product of the present disclosure can be written in any combination of one or more programming languages, and the program code can be executed completely on the user device, partially on the user device, as an independent software package, partially on the user device and partially on a remote device, or completely on the remote device.
[0243] Although the specific embodiments of the present disclosure are described above, those skilled in the art should understand that this is only an example, and the protection scope of the present disclosure is defined by the appended claims. Those skilled in the art may make various changes or modifications to these embodiments without departing from the principles and essence of the present disclosure, but these changes and modifications all fall within the protection scope of the present disclosure.
Claims
1. A method for detecting abnormal sound, It is characterized in that Applied to an active noise reduction system, the active noise reduction system includes a microphone and a speaker; the abnormal sound detection method includes: Acquire a first noise signal during the operation of the active noise reduction system; the first noise signal includes a noise signal of a noise source collected by the microphone and / or a control signal of the speaker; The first noise signal is subjected to Fourier transform at a target frequency resolution, and stable single-frequency abnormal sound detection is performed on the active noise reduction system according to the Fourier transform result; wherein the target frequency resolution is obtained by optimization, and the optimization goal is to accurately identify the abnormal sound and minimize the calculation amount of the abnormal sound detection; the characteristics of the stable single-frequency abnormal sound are: there is a prominent peak on the noise spectrum, the amplitude of the highest point of the peak is higher than the amplitude corresponding to the adjacent frequency, and the power of the peak accounts for a proportion of the total noise power of the first noise signal that is less than or equal to the power threshold.
2. The abnormal sound detection method according to claim 1, It is characterized in that The step of determining the target frequency resolution comprises: Acquire an abnormal sample; the abnormal sample is a noise signal containing an abnormal sound; The frequency resolution is iteratively updated, and the frequency resolution corresponding to the time when the abnormal sound of the abnormal sample is detected and the calculation amount is minimized is determined as the target frequency resolution.
3. The abnormal sound detection method according to claim 1, It is characterized in that Acquiring a first noise signal during the operation of the active noise reduction system includes: A first noise signal during the operation of the active noise reduction system is obtained at a target sampling frequency; the target sampling frequency is determined according to an abnormal sound frequency and / or a computing power of an analysis device of the first noise signal; and the abnormal sound frequency is determined according to a noise sample containing an abnormal sound.
4. The abnormal sound detection method according to claim 3, It is characterized in that The target sampling frequency is greater than or equal to 2 times the out-of-tone frequency; And / or, the computing power is characterized by a maximum number of data points for each Fourier transform performed by the analysis device, and the target sampling frequency is less than or equal to the product of the maximum number of data points and the frequency of the Fourier transform.
5. The abnormal sound detection method according to claim 1, It is characterized in that Performing abnormal sound detection on the active noise reduction system according to the Fourier transform result includes: Determine the total amplitude of the spectral power of the first noise signal according to the Fourier transform result; In response to the total amplitude of the spectrum power being within a preset power range, it is determined that the active noise reduction system has a stable single-frequency abnormal sound; wherein the lower limit of the preset power range is greater than the maximum power of the noise signal of the noise source, and the upper limit of the preset power range is less than the power of the first noise signal when howling occurs in the active noise reduction system; and / or, in response to the difference between the total amplitude of the spectrum power and the noise peak being less than a first difference threshold, determining that a stable single-frequency abnormal sound exists in the active noise reduction system; And / or, in response to the presence of a noise segment in the first noise signal whose power ratio is greater than or equal to a ratio threshold, determining that a stable single-frequency abnormal sound exists in the active noise reduction system; the power ratio is the ratio of the power of the noise segment to the total amplitude of the spectrum power; And / or, in response to the difference between the total amplitude of the spectrum power and the filter power being greater than or equal to a second difference threshold, determining that the active noise reduction system has a stable single-frequency abnormal sound; the filter power is the result of filtering out the noise peak from the total amplitude of the spectrum power; And / or, in response to the presence of a signal segment having a power difference with an adjacent signal segment greater than a third difference threshold, it is determined that a stable single-frequency alien sound exists.
6. The abnormal sound detection method according to claim 5, It is characterized in that The preset power range is determined by: Acquire abnormal samples and normal samples; the abnormal samples are noise signals containing abnormal sounds, and the normal samples are noise signals not containing abnormal sounds; Constructing an objective function; the objective function is constructed according to at least one of the following parameters: the number of missed detections of abnormal samples, the number of false detections of normal samples, the percentage of missed detections of abnormal samples, and the percentage of false detections of normal samples; Optimizing the objective function within a first preset optimization range of the upper limit value and / or the lower limit value to determine a final preset power range; And / or, the difference threshold is determined by: Acquire abnormal samples and normal samples; the abnormal samples are noise signals containing abnormal sounds, and the normal samples are noise signals not containing abnormal sounds; Constructing an objective function; the objective function is constructed according to at least one of the following parameters: the number of missed detections of abnormal samples, the number of false detections of normal samples, the percentage of missed detections of abnormal samples, and the percentage of false detections of normal samples; The objective function is optimized within a second preset optimization range of the difference threshold to determine a final difference threshold.
7. The abnormal sound detection method according to claim 6, It is characterized in that The first preset optimization range is [74dB, 91dB]; And / or, the second preset optimization range is [1dB, 10dB]; And / or, the optimization algorithm includes at least one of the following algorithms: genetic algorithm, simulated annealing algorithm, particle swarm algorithm, Bayesian optimization algorithm, hill climbing algorithm; And / or, the objective function is the ratio of the number of anomalies to the total number of samples, and the number of anomalies is the sum of the number of missed detections of the abnormal samples and the number of false detections of normal samples.
8. The abnormal sound detection method according to any one of claims 1 to 7, It is characterized in that Before obtaining the first noise signal during the operation of the active noise reduction system, the method further includes: When the active noise reduction system is not started, obtaining a second noise signal from a noise source; Performing segmentation processing on the second noise signal so that the power proportion of each noise segment of the second noise signal is less than a proportion threshold, and determining a segmentation strategy for the second noise signal; Performing Fourier transform on the first noise signal includes: Segment-process the first noise signal according to the segmentation strategy, and perform Fourier transform on the noise segments obtained by segmentation; Alternatively, before obtaining the first noise signal during the operation of the active noise reduction system, the method further includes: When the active noise reduction system is not started, obtaining a second noise signal from a noise source; performing octave calculation on the second noise signal to determine a segmentation strategy matching the second noise signal; Segmenting the second noise signal according to the segmentation strategy, and calculating the characteristic value of each noise segment of the second noise signal; the characteristic value includes power and / or decibel value; Determine a target gain of a first target noise segment whose characteristic value is greater than a characteristic value threshold; after filtering the first target noise segment with a bandpass filter of the target gain, the characteristic value of the first target noise segment is less than or equal to the characteristic value threshold; Performing Fourier transform on the first noise signal includes: The first noise signal is segmented to obtain at least two noise segments, and a second target noise segment contained in the first noise signal is filtered using a bandpass filter with a target gain, and an unfiltered noise segment and a filtered noise segment in the first noise signal are respectively Fourier transformed; wherein a frequency band of the second target noise segment includes a frequency band of the first target noise segment.
9. A device for detecting abnormal sound, It is characterized in that Applied to an active noise reduction system, the active noise reduction system comprising a microphone and a speaker; the abnormal sound detection device is used to implement the abnormal sound detection method described in any one of claims 1 to 8; The abnormal sound detection method comprises: An acquisition module, used to acquire a first noise signal during the operation of the active noise reduction system; the first noise signal is a noise signal of a noise source collected by the microphone and / or a control signal of the speaker; A detection module is used to perform Fourier transform on the first noise signal with a target frequency resolution, and perform stable single-frequency abnormal sound detection on the active noise reduction system according to the Fourier transform result; wherein the target frequency resolution is obtained by optimization, and the optimization goal is to accurately identify the abnormal sound and minimize the calculation amount of the abnormal sound detection; the characteristics of the stable single-frequency abnormal sound are: there is a prominent peak on the noise spectrum, the amplitude of the highest point of the peak is higher than the amplitude corresponding to the adjacent frequency, and the power of the peak accounts for a proportion of the total noise power of the first noise signal that is less than or equal to the power threshold.
10. An active noise reduction system comprising a memory, a processor, and a computer program stored in the memory and running on the processor, It is characterized in that When the processor executes the computer program, the method according to any one of claims 1 to 8 is implemented.
Citation Information
Patent Citations
Power cable field partial discharge signal noise reduction method
CN107728018A
Noise estimation method and system
CN109741760A
Active noise reduction method and device and audio playing equipment
CN113973248A
Abnormal sound detection method, computer readable storage medium and electronic equipment
CN114974301A
Measurement method for perceiving audio quality
CN117676446A
Cited By
Active noise reduction method, system and device
CN120032619A
