A split and detachable operating device with audio data collection function

Abstract

The present invention relates to the technical field of data denoising, and specifically to a split and detachable operating device with an audio data acquisition function; obtaining a peak clustering cluster according to the peak point of an audio data waveform; determining an audio noise waveform according to the interval characteristics and distribution characteristics of the peak points in the peak clustering cluster and the data difference characteristics of the peak clustering cluster; obtaining a mutation degree according to the data difference characteristics of the peak point of a preset low-frequency part and the adjacent trough point in the spectrum diagram of the audio noise waveform; obtaining a mixed feature point according to the mutation degree and interval characteristics of the peak point; obtaining a non-noise frequency point according to the difference characteristics of the amplitude of the mixed feature point and the same frequency of the adjacent audio acquisition port. The present invention filters out the preset low-frequency part in the spectrum diagram and retains the non-noise frequency point, obtains an adaptive filtering spectrum diagram, and obtains an audio denoising waveform diagram according to the adaptive filtering spectrum diagram; retains the human voice during denoising to improve the audio quality.

Description

Technical Field

[0001] The present invention relates to the technical field of data denoising, and in particular to a split and detachable operating device with an audio data acquisition function. Background Art

[0002] With the continuous development of the construction and manufacturing industries, the scale of construction and production is growing. In order to visualize the progress of work and ensure the safety of employees, audio and video equipment is gradually applied to the construction and production links. Through audio and video equipment, the progress of work and the status of employees can be monitored. In order to make the audio and video equipment portable and easy to disassemble, it can be designed as a split and detachable style so that it can be installed on a safety helmet. The split and detachable working device with audio acquisition function is often used in large-scale construction and production sites. The audio data obtained in this scene is prone to contain noise such as wind noise, which affects the audio quality. Therefore, in order to reduce the interference of wind noise, the frequency domain filtering method can be used for denoising according to the relatively low-frequency characteristics of wind noise; but since the male voice also contains low-frequency data, if low-frequency filtering is used directly, it may cause the loss of non-noise audio, making the audio incomplete, affecting the quality of audio data acquisition. Summary of the invention

[0003] In order to solve the technical problem that the above-mentioned method of using frequency domain filtering to denoise audio data leads to incomplete audio and affects the quality of audio data collection, the purpose of the present invention is to provide a split and detachable working device with audio data collection function, and the technical solution adopted is as follows:

[0004] A data acquisition module, used to obtain an audio data waveform;

[0005] A noise analysis module, for clustering the peak points of the audio data waveform to obtain different peak clustering clusters; obtaining the degree of noise interference according to the interval characteristics and distribution characteristics of the peak points in the peak clustering clusters and the data difference characteristics of different peak clustering clusters; and determining the audio noise waveform that needs to be denoised according to the noise interference degree;

[0006] A noise determination module is used to obtain a spectrum diagram of an audio noise waveform diagram that needs to be denoised; obtain the mutation degree of the peak point according to the data difference characteristics of the peak point and the adjacent trough point of the preset low-frequency part in the spectrum diagram; obtain the mixed fluctuation degree of the peak point according to the mutation degree of the peak point, the mutation degree and interval characteristics of the adjacent peak points of the peak point; obtain the mixed feature point according to the mixed fluctuation degree; obtain the non-noise frequency point according to the difference characteristics of the amplitude of the mixed feature point and the same frequency of the adjacent audio acquisition port;

[0007] The audio denoising module is used to filter out the preset low-frequency part in the spectrum diagram and retain the non-noise frequency points to obtain an adaptive filtering spectrum diagram; and obtain an audio denoising waveform diagram according to the adaptive filtering spectrum diagram.

[0008] Furthermore, the step of obtaining the degree of noise interference according to the interval characteristics and distribution characteristics of the peak points in the peak clustering clusters and the data difference characteristics of different peak clustering clusters includes:

[0009] Calculate the average value of the time span of the peak points in each peak clustering cluster and negatively map them to obtain the cluster width eigenvalue; calculate the sum of the numerical variances of the peak points in each peak clustering cluster and normalize them to obtain the audio fluctuation eigenvalue; calculate the absolute value of the difference between the average values ​​of the peak points of the peak clustering cluster and the adjacent peak clustering clusters to obtain the peak difference value of the peak clustering cluster; calculate the average value of the peak difference values ​​of the peak clustering clusters and negatively map them to obtain the peak similarity eigenvalue; calculate the sum of the cluster width eigenvalue, the audio fluctuation eigenvalue and the peak similarity eigenvalue to obtain the noise interference degree.

[0010] Furthermore, the step of determining the audio data waveform graph that needs to be denoised according to the noise interference degree includes:

[0011] When the noise interference degree is greater than a preset interference threshold, the audio data waveform is used as an audio noise waveform that needs to be denoised.

[0012] Furthermore, the step of obtaining the mutation degree of the peak point according to the data difference characteristics of the peak point and the adjacent trough point of the preset low-frequency part in the spectrum graph includes:

[0013] The amplitude difference between the peak point and the adjacent trough point is calculated to obtain the amplitude change value of the peak point; and the sum of the two amplitude change values ​​of the peak point is calculated to obtain the mutation degree of the peak point.

[0014] Furthermore, the step of obtaining the mixed fluctuation degree of the crest point according to the mutation degree of the crest point, the mutation degree of adjacent crest points of the crest point and the interval characteristics comprises:

[0015] The time interval between the peak point and the adjacent peak point is used as the interval length value; the ratio of the mutation degree of the adjacent peak points of the peak point to the interval length value is calculated to obtain the abnormal characteristic value of the adjacent peak points of the peak point; the sum of the abnormal characteristic values ​​of all adjacent peak points of the peak point is calculated and normalized to obtain the first mixed characteristic value; the mutation degree of the peak point is normalized to obtain the second mixed characteristic value; the sum of the first mixed characteristic value and the second mixed characteristic value is calculated to obtain the mixed fluctuation degree of the peak point.

[0016] Furthermore, the step of obtaining the mixed feature points according to the mixed fluctuation degree includes:

[0017] When the mixed fluctuation degree of the peak point exceeds a preset mixed fluctuation threshold, the peak point is a mixed characteristic point.

[0018] Furthermore, the step of acquiring the non-noise frequency point according to the difference feature between the mixed feature point and the amplitude of the same frequency of the adjacent audio acquisition port includes:

[0019] The amplitude of the same frequency as the mixed feature point in the spectrum diagram of the adjacent audio acquisition port is used as the reference amplitude; the absolute value of the difference between the amplitude of the mixed feature point and the reference amplitude is calculated and negatively correlated to obtain the noise resolution value of the mixed feature point; when the noise resolution value exceeds a preset resolution threshold, the mixed feature point is a non-noise frequency point.

[0020] Furthermore, the step of clustering according to the peak points of the audio data waveform to obtain different peak clustering clusters includes:

[0021] The peak points in the audio data waveform are clustered using a K-means clustering algorithm to obtain different peak clustering clusters.

[0022] The present invention has the following beneficial effects:

[0023] In the present invention, obtaining peak clustering clusters can be used to analyze whether the audio data waveform contains noise, providing a basis for more accurate denoising; obtaining the degree of noise interference can be analyzed according to the difference in the waveform change distribution of human voice and wind noise in the audio data waveform, so as to obtain the audio noise waveform with noise; avoiding denoising the audio data that does not need to be denoised, resulting in the elimination of part of the human voice affecting the audio quality, and preliminarily improving the denoising effect. Obtaining the spectrum of the audio noise waveform can perform frequency domain filtering according to the low-frequency characteristics of wind noise, and filter out the wind noise to the greatest extent; obtaining the mutation degree of the peak point can determine the frequency of the low-frequency part of the human voice and wind noise in the spectrum; obtaining the mixed fluctuation degree can further determine the mixed feature points that may be the human voice according to the position characteristics and mutation degree of the peak point and the adjacent peak point. Obtaining non-noise frequency points can more accurately determine the frequency points where the human voice appears according to the difference characteristics of wind noise at different audio acquisition ports, so as to retain them when filtering the low-frequency data. Acquiring the adaptive filtering spectrum diagram can preserve the low-frequency human voice in the process of wind noise removal, making denoising more accurate; finally obtaining the audio denoising waveform diagram can improve the audio quality of the split detachable device. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0025] Figure 1 A block diagram of a split and detachable operating device with an audio data acquisition function provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0026] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of the specific implementation, structure, features and effects of a split and detachable operating device with an audio data acquisition function proposed by the present invention in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0027] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0028] The specific scheme of a split and detachable operating device with an audio data acquisition function provided by the present invention is described in detail below with reference to the accompanying drawings.

[0029] See also Figure 1 , which shows a block diagram of a split and detachable operating device with an audio data acquisition function provided by an embodiment of the present invention, the device includes the following modules:

[0030] The data acquisition module S1 is used to acquire the audio data waveform.

[0031] In an embodiment of the present invention, the implementation scenario is to denoise the audio data of a split detachable device with an audio data acquisition function to improve the quality of the collected audio. First, an audio data waveform is obtained. The split detachable device is installed on a safety helmet. There is an audio acquisition port on both sides of the helmet. The audio data received by each port constructs an audio data waveform. In an embodiment of the present invention, the acquisition frequency is 44.1KHz. The horizontal axis of the audio data waveform is the acquisition time, and the vertical axis is the amplitude. A segment of the audio data waveform is obtained every 10 seconds. The implementer can set the acquisition parameters according to the implementation scenario.

[0032] The noise analysis module S2 is used to cluster the peak points of the audio data waveform to obtain different peak clustering clusters; obtain the degree of noise interference based on the interval characteristics and distribution characteristics of the peak points in the peak clustering clusters and the data difference characteristics of different peak clustering clusters; and determine the audio noise waveform that needs to be denoised based on the degree of noise interference.

[0033] Since the device is often used in large-scale buildings and production sites, it is easily disturbed by wind noise when recording the user's audio, which affects the audio quality; therefore, the audio needs to be denoised to improve the audio quality. Wind noise usually presents low-frequency characteristics, so the low-frequency data in the audio data can be filtered out, but since the human voice includes some low-frequency parts, the low-frequency data that needs to be denoised needs to be further analyzed during denoising to improve the audio quality. In order to avoid denoising the audio data without noise, resulting in excessive denoising affecting the audio quality, it is necessary to analyze whether the audio segment is affected by wind noise before denoising. If there is no wind noise, it can be directly transmitted and stored, reducing the denoising processing time and reducing the situation of excessive denoising. When the user is speaking, the vibration generated will cause a large fluctuation in the audio data, and the pauses between pronunciations will cause the waveform distribution in the audio data waveform to present block intervals, and there will be obvious intervals between the high-amplitude waveform and the low-amplitude waveform. When wind noise occurs, the continuous wind flow passing through the audio acquisition port will cause a continuous waveform distribution in the audio data waveform. There is no obvious boundary between the high-amplitude waveform and the low-amplitude waveform. The wind noise causes many random subtle peaks to appear in the waveform. Therefore, firstly, clustering is performed according to the peak points of the audio data waveform to obtain different peak clustering clusters; preferably, in an embodiment of the present invention, the peak points in the audio data waveform are clustered by the K-means clustering algorithm to obtain different peak clustering clusters. It should be noted that the K-means clustering algorithm belongs to the prior art and the specific steps will not be repeated. The value of the peak point in each peak cluster is close to the time.

[0034] Furthermore, the degree of noise interference can be obtained based on the interval characteristics and distribution characteristics of the peak points in the peak clustering clusters, and the data difference characteristics of different peak clustering clusters; preferably, in an embodiment of the present invention, the step of obtaining the degree of noise interference includes: calculating the average value of the time span of the peak points in each peak clustering cluster and negatively mapping it to obtain a cluster width characteristic value; when the time span of the peak points in the peak clustering cluster is longer, the cluster width characteristic value is smaller, which means that the distribution of the peak clustering cluster in the waveform graph is wider, and it is more consistent with the characteristics of the existence of block intervals in the waveform when the user speaks, the possibility of random peak points caused by wind noise is lower, and the possibility of small peak clustering clusters is lower; when the cluster width characteristic value is larger, it means that the time span of the peak points in each peak clustering cluster is smaller, the more random peak points caused by wind noise, and the more small peak clustering clusters are. The sum of the numerical variances of the peak points in each peak cluster is calculated and normalized to obtain the audio fluctuation characteristic value; when the numerical variance of the peak points in the peak cluster is larger, it means that the numerical difference of the peak points in the peak cluster is larger, and the audio fluctuation characteristic value is larger, which means that it is more consistent with the situation of numerical instability of random peak points caused by wind noise, and less consistent with the situation that the peak point values ​​are relatively stable when the user speaks. The absolute value of the difference between the average values ​​of the peak points of the peak cluster and the adjacent peak cluster is calculated to obtain the peak difference value of the peak cluster; in the embodiment of the present invention, the absolute value of the difference between the average values ​​of the peak points of the peak cluster and the next adjacent peak cluster is calculated, and the last peak cluster does not participate in the calculation. The larger the peak difference value, the greater the numerical difference of the peak points of the two peak clusters, the greater the waveform difference, and the more consistent with the difference characteristics of the high-amplitude waveform and the low-amplitude waveform. Calculate the average value of the peak difference value of the peak clustering cluster and negatively map it to obtain the peak similarity eigenvalue; when the average value of the peak difference value is smaller, the peak similarity eigenvalue is larger, which means that the difference between the high-amplitude waveform and the low-amplitude waveform in the waveform graph is smaller, and the number of random peak points caused by wind noise is more, making the difference between the high-amplitude waveform and the low-amplitude waveform not obvious. Calculate the sum of the cluster width eigenvalue, the audio fluctuation eigenvalue and the peak similarity eigenvalue to obtain the degree of noise interference; when the degree of noise interference is greater, it means that the audio data is more affected by wind noise, the audio quality is lower, and the more denoising is needed. The formulas for obtaining the degree of noise interference include:

[0035]

[0036] Where E represents the noise interference degree of the audio data waveform. represents an exponential function with a natural constant as the base, K represents the number of peak clusters, represents the time span of the peak point of the kth peak cluster, represents the cluster width eigenvalue, Represents the numerical variance of the peak point of the kth peak cluster, represents the normalization function, Represents the audio fluctuation characteristic value, represents the peak difference value of the kth peak cluster, Represents the peak similarity eigenvalue.

[0037] The greater the noise interference level of the audio data waveform, the more denoising is needed. Therefore, the audio noise waveform that needs to be denoised is determined according to the noise interference level. Preferably, in an embodiment of the present invention, the step of determining the audio noise waveform that needs to be denoised includes: when the noise interference level is greater than a preset interference threshold, the audio data waveform is used as the audio noise waveform that needs to be denoised. In an embodiment of the present invention, the preset interference threshold is 1.2, which can be determined by the implementer according to the implementation scenario; when the noise interference level of the audio data waveform does not exceed the preset interference threshold, it means that the wind noise is small and has a low impact on the audio quality, and denoising processing may not be performed, thereby avoiding excessive denoising that reduces the audio quality; for the audio noise waveform that needs to be denoised, accurate denoising analysis is required.

[0038] The noise determination module S3 is used to obtain the spectrum of the audio noise waveform that needs to be denoised; obtain the mutation degree of the peak point according to the data difference characteristics of the peak point and the adjacent trough points in the preset low-frequency part of the spectrum; obtain the mixed fluctuation degree of the peak point according to the mutation degree of the peak point, the mutation degree of the adjacent peak points of the peak point and the interval characteristics; obtain the mixed feature point according to the mixed fluctuation degree; obtain the non-noise frequency point according to the difference characteristics of the amplitude of the mixed feature point and the same frequency of the adjacent audio acquisition port.

[0039] Wind noise usually has low-frequency characteristics, so the relatively low-frequency part of the waveform can be filtered out, so the spectrum of the audio noise waveform that needs to be denoised is obtained. The spectrum can be obtained through the existing Fourier transform. The specific steps are not repeated. The horizontal axis of the spectrum is the frequency, and the vertical axis is the amplitude of different frequency signals. The wind noise is located in the low-frequency part of the spectrum. Since the human voice of the user also contains a part of the low-frequency part, if the low-frequency part is directly filtered out, the audio quality will be further reduced. Therefore, it is necessary to further analyze according to the characteristics of the spectrum and perform adaptive filtering. The distribution of wind noise in the low frequency has a certain trend law. The higher the frequency of wind noise, the smaller the amplitude intensity corresponding to it; as the frequency increases, the corresponding amplitude intensity is smaller, and the amplitude curve of wind noise shows a downward trend, and there are slight fluctuations in the process of decline. When there are continuous peaks with strong mutations in the low-frequency part of the spectrum that do not conform to the characteristics of the wind noise spectrum, it means that some of the peaks in the low-frequency part are caused by wind noise mixed with the low-frequency part of the human voice. Such peak points need to be retained to avoid filtering out some human voices and reducing the audio quality. Therefore, the mutation degree of the peak point is obtained according to the data difference characteristics of the peak point and the adjacent trough point of the preset low-frequency part in the spectrum diagram. In the embodiment of the present invention, the preset low-frequency part is 0-3KHz, and the implementer can determine it according to the implementation scenario.

[0040] Preferably, in an embodiment of the present invention, the step of obtaining the degree of mutation includes: calculating the amplitude difference between the peak point and the adjacent trough point to obtain the amplitude change value of the peak point; the adjacent trough point is the nearest trough point on the left and right of the peak point, and if the peak point is located at the beginning and end points, it does not participate in the calculation. Calculate the sum of the two amplitude change values ​​of the peak point to obtain the degree of mutation of the peak point; when the amplitude change value is larger, it means that the degree of mutation at the peak point is larger, and it is more likely to be a feature of the mixture of human voice and wind noise. If there are similar characteristic peak points around the peak point, it can be more likely that the peak point reflects the characteristics of the mixture of human voice and wind noise; therefore, the mixed fluctuation degree of the peak point can be obtained according to the degree of mutation of the peak point, the degree of mutation of the adjacent peak points of the peak point, and the interval characteristics; preferably, in an embodiment of the present invention, the step of obtaining the degree of mixed fluctuation includes: taking the time interval between the peak point and the adjacent peak point as the interval length value; the adjacent peak point is the nearest peak point on the left and right of the peak point. Calculate the ratio of the mutation degree of the adjacent peak points of the peak point to the interval length value to obtain the abnormal characteristic value of the adjacent peak points of the peak point; when the mutation degree of the adjacent peak points of the peak point is greater and the distance to the peak point is closer, the abnormal characteristic value is greater, and the peak point is more likely to reflect the characteristics of the mixture of human voice and wind noise. Calculate the sum of the abnormal characteristic values ​​of all adjacent peak points of the peak point and normalize them to obtain the first mixed characteristic value; normalize the mutation degree of the peak point to obtain the second mixed characteristic value; calculate the sum of the first mixed characteristic value and the second mixed characteristic value to obtain the mixed fluctuation degree of the peak point; when the mixed fluctuation degree is greater, it means that the peak point is more likely to be generated by the mixture of wind noise and human voice.

[0041] Furthermore, a mixed feature point can be obtained according to the mixed fluctuation degree, specifically including: when the mixed fluctuation degree of the peak point exceeds the preset mixed fluctuation threshold, the peak point is a mixed feature point; in the embodiment of the present invention, the preset mixed fluctuation threshold is 0.6, which can be determined by the implementer according to the implementation scenario. Due to the randomness of wind in the environment, a peak with a strong mutation degree may be generated in the low-frequency area when the wind changes; therefore, the currently obtained mixed feature points may have peaks caused by wind noise only, so it is necessary to further analyze the mixed feature points to improve the accuracy of denoising.

[0042] The split detachable device in the embodiment of the present invention has two audio acquisition ports, which are located on both sides of the safety helmet. Since the degree of flow of the wind flow in each place is different, the degree of interference of wind noise on each port is different. The wind noise in the audio data collected in each port is different, but the audio data is the same when the human voice is transmitted to the ports on both sides. Therefore, in the spectrum diagram corresponding to the audio data waveform diagram of the same period of time of the two ports, the peak points with higher similarity in the low-frequency part are more likely to be caused by the transmission of human voice, while the peak points with lower similarity are more likely to be caused by wind noise. Therefore, the non-noise frequency point can be obtained according to the difference characteristics of the amplitude of the same frequency of the mixed feature point and the adjacent audio acquisition port; preferably, in the embodiment of the present invention, the step of obtaining the non-noise frequency point includes: taking the amplitude of the same frequency as the mixed feature point in the spectrum diagram of the adjacent audio acquisition port as the reference amplitude; calculating the absolute value of the difference between the amplitude of the mixed feature point and the reference amplitude and negatively mapping them to obtain the noise resolution value of the mixed feature point; when the difference between the two amplitudes is greater, the noise resolution value is smaller, and the mixed feature point is more likely to be caused by wind noise, and when the difference between the two amplitudes is smaller, the noise resolution value is larger, and the point is more likely to be caused by human voice. When the noise resolution value exceeds the preset resolution threshold, the mixed feature point is a non-noise frequency point, and the non-noise frequency point characterizes that the amplitude at this frequency is more caused by human voice. In the embodiment of the present invention, the preset resolution threshold is 0.5, and the implementer can determine it according to the implementation scenario.

[0043] The audio denoising module S4 is used to filter out the preset low-frequency part in the spectrum diagram and retain the non-noise frequency points to obtain an adaptive filtering spectrum diagram; and obtain an audio denoising waveform diagram according to the adaptive filtering spectrum diagram.

[0044] After obtaining the non-noise frequency points in the preset low-frequency part, the preset low-frequency part can be filtered out in the spectrum diagram and the non-noise frequency points can be retained to obtain an adaptive filtering spectrum diagram; in the embodiment of the present invention, an existing high-pass filter is used for filtering, and the specific steps are not repeated; the non-noise frequency points are retained during the filtering process, so that the low-frequency human voice can be retained during the wind noise removal process, thereby improving the audio quality. After obtaining the adaptive filtering spectrum diagram, an audio denoising waveform diagram can be obtained according to the adaptive filtering spectrum diagram, and the audio denoising waveform diagram is obtained using the existing inverse Fourier transform, and the specific steps are not repeated. The audio denoising waveform diagram can remove the low-frequency wind noise while retaining the human voice part in the low frequency, ultimately improving the audio quality collected by the device; the audio denoising waveform diagram is stored or transmitted for subsequent use.

[0045] In summary, the embodiments of the present invention provide a split and detachable operating device with an audio data acquisition function; obtain peak clustering clusters according to the peak points of the audio data waveform; determine the audio noise waveform according to the interval characteristics and distribution characteristics of the peak points in the peak clustering clusters and the data difference characteristics of the peak clustering clusters; obtain the degree of mutation according to the data difference characteristics of the peak points and adjacent trough points of the preset low-frequency part in the spectrum diagram of the audio noise waveform; obtain mixed feature points according to the degree of mutation and interval characteristics of the peak points; obtain non-noise frequency points according to the difference characteristics of the amplitude of the mixed feature points and the same frequency of the adjacent audio acquisition ports. The present invention filters out the preset low-frequency part in the spectrum diagram and retains the non-noise frequency points to obtain an adaptive filtering spectrum diagram, and obtains an audio denoising waveform diagram according to the adaptive filtering spectrum diagram; retains the human voice during denoising to improve the audio quality.

[0046] It should be noted that the sequence of the above embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0047] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

Claims

1. A split and detachable working device with audio data collection function, characterized in that: The device comprises the following modules: A data acquisition module, used to obtain an audio data waveform; A noise analysis module, for clustering the peak points of the audio data waveform to obtain different peak clustering clusters; obtaining the degree of noise interference according to the interval characteristics and distribution characteristics of the peak points in the peak clustering clusters and the data difference characteristics of different peak clustering clusters; and determining the audio noise waveform that needs to be denoised according to the noise interference degree; A noise determination module, used to obtain a frequency spectrum diagram of an audio noise waveform diagram that needs to be denoised; Obtaining the mutation degree of the peak point according to the data difference characteristics of the peak point and the adjacent trough point in the preset low-frequency part of the spectrum graph; The mixed fluctuation degree of the peak point is obtained according to the mutation degree of the peak point, the mutation degree of the adjacent peak points of the peak point and the interval characteristics; the mixed characteristic point is obtained according to the mixed fluctuation degree; the non-noise frequency point is obtained according to the difference characteristics of the amplitude of the same frequency between the mixed characteristic point and the adjacent audio acquisition port; An audio denoising module is used to filter out a preset low-frequency part in the frequency spectrum and retain the non-noise frequency points to obtain an adaptive filtering frequency spectrum; An audio denoising waveform diagram is obtained according to the adaptive filtering spectrum diagram.

2. The split and detachable working device with audio data collection function according to claim 1 is characterized in that: The step of obtaining the degree of noise interference according to the interval characteristics and distribution characteristics of the peak points in the peak clustering clusters and the data difference characteristics of different peak clustering clusters includes: Calculate the average value of the time span of the peak points in each peak clustering cluster and negatively map them to obtain the cluster width eigenvalue; calculate the sum of the numerical variances of the peak points in each peak clustering cluster and normalize them to obtain the audio fluctuation eigenvalue; calculate the absolute value of the difference between the average values ​​of the peak points of the peak clustering cluster and the adjacent peak clustering clusters to obtain the peak difference value of the peak clustering cluster; calculate the average value of the peak difference values ​​of the peak clustering clusters and negatively map them to obtain the peak similarity eigenvalue; calculate the sum of the cluster width eigenvalue, the audio fluctuation eigenvalue and the peak similarity eigenvalue to obtain the noise interference degree.

3. The split and detachable working device with audio data collection function according to claim 1 is characterized in that: The step of determining the audio data waveform graph that needs to be denoised according to the noise interference degree comprises: When the noise interference degree is greater than a preset interference threshold, the audio data waveform is used as an audio noise waveform that needs to be denoised.

4. The split and detachable working device with audio data collection function according to claim 1 is characterized in that: The step of obtaining the mutation degree of the peak point according to the data difference characteristics between the peak point and the adjacent trough point of the preset low-frequency part of the spectrum graph comprises: The amplitude difference between the peak point and the adjacent trough point is calculated to obtain the amplitude change value of the peak point; and the sum of the two amplitude change values ​​of the peak point is calculated to obtain the mutation degree of the peak point.

5. The split and detachable working device with audio data collection function according to claim 1 is characterized in that: The step of obtaining the mixed fluctuation degree of the crest point according to the mutation degree of the crest point, the mutation degree of adjacent crest points of the crest point and the interval characteristics comprises: The time interval between the peak point and the adjacent peak point is used as the interval length value; the ratio of the mutation degree of the adjacent peak points of the peak point to the interval length value is calculated to obtain the abnormal characteristic value of the adjacent peak points of the peak point; the sum of the abnormal characteristic values ​​of all adjacent peak points of the peak point is calculated and normalized to obtain the first mixed characteristic value; the mutation degree of the peak point is normalized to obtain the second mixed characteristic value; the sum of the first mixed characteristic value and the second mixed characteristic value is calculated to obtain the mixed fluctuation degree of the peak point.

6. The split and detachable working device with audio data collection function according to claim 1 is characterized in that: The step of obtaining mixed feature points according to the mixed fluctuation degree comprises: When the mixed fluctuation degree of the peak point exceeds a preset mixed fluctuation threshold, the peak point is a mixed characteristic point.

7. The split and detachable working device with audio data collection function according to claim 1 is characterized in that: The step of acquiring the non-noise frequency point according to the difference feature between the mixed feature point and the amplitude of the same frequency of the adjacent audio acquisition port comprises: The amplitude of the same frequency as the mixed feature point in the spectrum diagram of the adjacent audio acquisition port is used as the reference amplitude; the absolute value of the difference between the amplitude of the mixed feature point and the reference amplitude is calculated and negatively correlated to obtain the noise resolution value of the mixed feature point; when the noise resolution value exceeds a preset resolution threshold, the mixed feature point is a non-noise frequency point.

8. The split and detachable working device with audio data collection function according to claim 1 is characterized in that: The step of clustering the peak points of the audio data waveform to obtain different peak clustering clusters comprises: The peak points in the audio data waveform are clustered using a K-means clustering algorithm to obtain different peak clustering clusters.

Images