Method, system and chip for adaptively adjusting noise reduction mode

Abstract

The present invention provides a method, system and chip for adaptively adjusting a noise reduction mode, including: extracting noise feature data from the sound information based on the sound information in the environment collected by a sound collection device; obtaining an initial noise reduction mode configured by the sound collection device; wherein the initial noise reduction mode includes a noise reduction algorithm and its corresponding initial noise reduction algorithm parameters; generating a first correction value based on the noise feature data; obtaining a sound collection method of the sound collection device, and generating a second correction value based on the sound collection method; performing a fusion analysis based on the first correction value and the second correction value to obtain a fusion correction value; and correcting the initial noise reduction algorithm parameters of the noise reduction algorithm based on the fusion correction value to obtain a corrected noise reduction algorithm for sound noise reduction processing. In the present invention, the noise reduction parameters of the noise reduction mode are adaptively adjusted to overcome the defect that the current noise reduction method cannot effectively remove background noise interference.

Description

Technical Field

[0001] The present invention relates to the technical field of speech processing, and in particular to a method, system and chip for adaptively adjusting a noise reduction mode. Background Art

[0002] Smartphones, smart headphones, professional recording devices and other devices have very different sound characteristics in different environments. The complex and changeable environmental noise makes it difficult for sound collection devices to collect high-quality audio.

[0003] Traditional sound collection devices usually use fixed noise reduction modes, and their noise reduction algorithm parameters are set when the device leaves the factory, or can only be switched through a limited number of preset modes, and cannot be flexibly and accurately adaptively adjusted according to the actual environmental noise conditions. This leads to poor noise reduction effects in some complex environments, excessive noise reduction will lose the authenticity of the sound, or insufficient noise reduction will fail to effectively remove background noise interference.

[0004] For example, in noisy streets, bustling restaurants, factory workshops and other environments, traditional sound collection devices have difficulty accurately identifying and processing different types of noise, such as continuous low-frequency noise, sudden high-frequency noise, and intermittent noise. Moreover, different sound collection methods, such as directional collection and omnidirectional collection, also have different requirements and processing methods for noise reduction, but traditional technologies fail to fully consider these factors for targeted optimization, resulting in unreasonable noise reduction methods that cannot effectively remove background noise interference. Summary of the invention

[0005] The main purpose of the present invention is to provide a method, system and chip for adaptively adjusting the noise reduction mode, aiming to overcome the defect that the current noise reduction method cannot effectively remove the background noise interference.

[0006] To achieve the above object, the present invention provides a method for adaptively adjusting a noise reduction mode, comprising the following steps:

[0007] The sound collection device collects sound information in the environment, and extracts noise characteristic data from the sound information;

[0008] Acquire an initial noise reduction mode configured by the sound collection device; wherein the initial noise reduction mode includes a noise reduction algorithm and its corresponding initial noise reduction algorithm parameters;

[0009] Based on the noise characteristic data, a first correction value is generated; a sound collection method of the sound collection device is obtained, and a second correction value is generated based on the sound collection method;

[0010] Perform fusion analysis based on the first correction value and the second correction value to obtain a fusion correction value;

[0011] The initial noise reduction algorithm parameters of the noise reduction algorithm are corrected based on the fusion correction value to obtain a corrected noise reduction algorithm for sound noise reduction processing.

[0012] Further, extracting noise characteristic data from the sound information includes:

[0013] Performing frequency analysis on the sound information, dividing it into different frequency ranges, and obtaining a plurality of frequency interval information;

[0014] Perform energy intensity calculation on each frequency interval information to obtain the energy intensity value of each frequency interval;

[0015] Perform noise determination processing based on the energy intensity value of each frequency interval to distinguish the noise frequency interval from the non-noise frequency interval;

[0016] The sound information in the noise frequency range is subjected to feature extraction processing to obtain noise feature data.

[0017] Further, generating a first correction value based on the noise characteristic data includes:

[0018] Performing spectrum analysis on the noise characteristic data to determine the main frequency distribution range of the noise;

[0019] Calculating the energy intensity mean of the noise within the main frequency distribution interval;

[0020] Compare the energy intensity mean with a plurality of preset energy intensity level thresholds to determine the energy intensity level to which the noise belongs;

[0021] According to the mapping relationship between the energy intensity level and the preset correction value, a corresponding first correction value is generated.

[0022] Further, generating a first correction value based on the noise characteristic data includes:

[0023] Performing spectrum analysis on the noise characteristic data to determine a main frequency distribution range of the noise, and generating a frequency difference value as a first sub-correction value according to a difference between the main frequency distribution range and a preset standard frequency distribution range;

[0024] Analyze the noise intensity information in the noise characteristic data, match it with a plurality of preset intensity levels, determine the intensity level to which it belongs, and generate an intensity correction value as a second sub-correction value according to a correction coefficient corresponding to the intensity level;

[0025] Extracting the noise duration feature in the noise feature data, dividing it into different time periods according to a preset time segmentation rule, counting the occurrence frequency of the noise in each time period, and generating a time frequency correction value as a third sub-correction value based on the occurrence frequency;

[0026] The first correction value is obtained by performing a weighted sum calculation on the first sub-correction value, the second sub-correction value and the third sub-correction value.

[0027] Furthermore, the sound collection method includes collection direction, collection distance setting parameters, and collection mode.

[0028] Further, generating a second correction value based on the sound collection method includes:

[0029] If the collection direction is directional collection, a direction deviation correction factor is generated as the first part of the correction value according to the deviation between the directional angle and the preset standard angle; if it is omnidirectional collection, an interference distribution correction factor is generated as the first part of the correction value based on the position distribution of other sound interference sources in the current environment relative to the sound collection device;

[0030] Comparing the collection distance setting parameters of the sound collection device with a preset optimal collection distance range, and generating a distance correction factor as a second part correction value according to the degree of deviation;

[0031] If the acquisition mode is continuous acquisition, a continuous acquisition correction factor is generated as the third part correction value according to the length of the acquisition time and the stability of the sound data; if the acquisition mode is intermittent acquisition, an intermittent acquisition correction factor is generated as the third part correction value according to the intermittent frequency and the duration of each acquisition;

[0032] The first correction value, the second correction value and the third correction value are fused and calculated to obtain the second correction value.

[0033] Further, the initial noise reduction algorithm parameters of the noise reduction algorithm are corrected based on the fusion correction value to obtain a corrected noise reduction algorithm for performing sound noise reduction processing, including:

[0034] The sound data obtained after the noise reduction process is stored in a voice database, and query permissions are configured for the sound data;

[0035] Performing hash operations on the first correction value and the second correction value to obtain corresponding first hash values ​​and second hash values;

[0036] Add the characters in the first hash value and the second hash value to the data table of the same specification one by one in sequence to obtain the corresponding first data table and second data table;

[0037] Performing change processing on the first data table and the second data table to obtain a third data table;

[0038] Based on the third data table and a preset algorithm, a query code is generated; wherein the query code is used to verify the query authority of the sound data.

[0039] Furthermore, the first data table and the second data table are changed to obtain a third data table, including:

[0040] Compare the cells at the same position of the first data table and the second data table one by one. When comparing, take the cells with the same character type at the same position of the first data table and the second data table as target cells, and delete all characters in other cells of the first data table and the second data table; if there are no cells with the same character type, take all cells as target cells;

[0041] For each target cell in the first data table, the characters in the target cell at the corresponding position in the second data table are combined with the characters in the target cell in the first data table to obtain the third data table.

[0042] Further, based on the third data table and the preset algorithm, a query code is generated, including:

[0043] Obtain the total number of blank cells in the third data table; based on the total number, generate a circle; wherein the radius of the circle^2=a+total number / 2, a is a preset value, and the column width and row height of the third data table are both one unit;

[0044] Overlaying the circle onto the third data table; wherein the center of the circle is located at the center of the third data table;

[0045] All characters in the circle are obtained and combined in sequence to obtain a combined string; the combined string is screened to obtain the query code.

[0046] The present invention also provides a system for adaptively adjusting a noise reduction mode, comprising:

[0047] An extraction unit, configured to extract noise characteristic data from the sound information collected in the environment by the sound collection device;

[0048] An acquisition unit, configured to acquire an initial noise reduction mode configured by the sound collection device; wherein the initial noise reduction mode includes a noise reduction algorithm and its corresponding initial noise reduction algorithm parameters;

[0049] A generating unit, configured to generate a first correction value based on the noise characteristic data; obtain a sound collection method of the sound collection device, and generate a second correction value based on the sound collection method;

[0050] a fusion unit, configured to perform fusion analysis based on the first correction value and the second correction value to obtain a fusion correction value;

[0051] The noise reduction unit is used to modify the initial noise reduction algorithm parameters of the noise reduction algorithm based on the fusion correction value to obtain a modified noise reduction algorithm for sound noise reduction processing.

[0052] The present invention also provides a computer chip, which is used to implement the steps of any of the above methods when executing a computer program.

[0053] The method, system and chip for adaptively adjusting the noise reduction mode provided by the present invention include: extracting noise feature data from the sound information in the sound collection environment based on the sound collection device; obtaining the initial noise reduction mode configured by the sound collection device; wherein the initial noise reduction mode includes a noise reduction algorithm and its corresponding initial noise reduction algorithm parameters; generating a first correction value based on the noise feature data; obtaining the sound collection method of the sound collection device, and generating a second correction value based on the sound collection method; performing fusion analysis based on the first correction value and the second correction value to obtain a fusion correction value; correcting the initial noise reduction algorithm parameters of the noise reduction algorithm based on the fusion correction value to obtain a corrected noise reduction algorithm for sound noise reduction processing. In the present invention, based on the characteristic data of the noise and the collection method of the sound collection device, the noise reduction parameters of the noise reduction mode are adaptively adjusted to achieve a better noise reduction effect, thereby overcoming the defect that the current noise reduction method is unreasonable and cannot effectively remove background noise interference. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 is a schematic diagram of the method steps for adaptively adjusting the noise reduction mode in one embodiment of the present invention;

[0055] Figure 2 It is a system structure block diagram of adaptively adjusting the noise reduction mode in one embodiment of the present invention.

[0056] The implementation, functional features and advantages of the present invention will be further described in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0057] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0058] Reference Figure 1 In one embodiment of the present invention, a method for adaptively adjusting a noise reduction mode is provided, comprising the following steps:

[0059] Step S1, collecting sound information in the environment based on a sound collection device, and extracting noise feature data from the sound information;

[0060] Step S2, obtaining an initial noise reduction mode configured by the sound collection device; wherein the initial noise reduction mode includes a noise reduction algorithm and its corresponding initial noise reduction algorithm parameters;

[0061] Step S3, generating a first correction value based on the noise characteristic data; obtaining a sound collection method of the sound collection device, and generating a second correction value based on the sound collection method;

[0062] Step S4, performing fusion analysis based on the first correction value and the second correction value to obtain a fusion correction value;

[0063] Step S5, correcting the initial noise reduction algorithm parameters of the noise reduction algorithm based on the fusion correction value to obtain a corrected noise reduction algorithm to perform sound noise reduction processing.

[0064] In this embodiment, it is intended to realize adaptive adjustment of the noise reduction mode according to the environment and the characteristics of the sound collection device itself, so as to improve the sound collection quality. It solves the problem that the noise reduction mode of the traditional sound collection device is fixed and cannot flexibly adapt to the complex and changeable environmental noise. In addition, it is difficult for traditional equipment to achieve good noise reduction effects in different environments, and it is easy to have excessive noise reduction or insufficient noise reduction. At the same time, it makes up for the fact that traditional technology does not fully consider the impact of differences in sound collection methods on noise reduction. Different collection methods (such as directional collection and omnidirectional collection) have different requirements and processing methods for noise reduction. This solution specifically generates corresponding correction values ​​according to the collection method to correct the noise reduction parameters.

[0065] Specifically, as described in step S1 above, the sound collection device collects the sound in the surrounding environment in real time through a built-in microphone or other audio sensor. The sound information collected above contains various sound components, including the required target sound and the noise in the environment. The collected sound information is analyzed and processed through an audio signal processing algorithm to extract the noise feature data therein. The noise feature data may include key feature information such as the frequency distribution, intensity, and duration of the noise. The above process provides basic data for subsequent noise reduction processing. Only by accurately extracting the noise features can noise reduction correction be carried out in a targeted manner.

[0066] As described in step S2 above, the sound collection device is configured with an initial noise reduction mode when it leaves the factory or after initialization. The above initial noise reduction mode includes a specific noise reduction algorithm and the corresponding initial noise reduction algorithm parameters. The noise reduction algorithm is a collection of a series of mathematical operations and logical judgments for processing noise, and the initial noise reduction algorithm parameters are the preset values ​​of these algorithms when they start running, such as the intensity of noise reduction, frequency range, etc. The above initial noise reduction mode provides a basic noise reduction processing framework, and the sound is preliminarily processed with noise reduction according to this initial mode before adaptive adjustment is performed.

[0067] As described in step S3 above, the extracted noise feature data is deeply analyzed. For example, according to the frequency distribution of the noise, if it is found that the noise is mainly concentrated in the low frequency band and is different from the common ambient noise frequency distribution model, then a frequency-related correction value is calculated based on the degree of difference. For the noise intensity, it is compared with the preset intensity level, and the intensity correction value is determined according to the level it is in. At the same time, the duration characteristics of the noise are analyzed, and the frequency of occurrence of the noise in different time periods is counted to generate a time-related correction value. Finally, these correction values ​​based on different noise characteristics are comprehensively calculated to obtain the first correction value. The first correction value mainly reflects the adjustment requirements of the noise reduction algorithm parameters due to the characteristics of the ambient noise itself, so that the noise reduction process can more accurately adapt to noises of different types and characteristics.

[0068] The collection method of the sound collection device will also affect the sound collection effect, so it is necessary to adjust the noise reduction effect in a targeted manner. For example, if the sound collection device adopts a directional collection method, it is necessary to consider the deviation between the collection direction and the ideal direction, as well as the positional relationship of the objects or noise sources in the surrounding environment that affect the collection effect relative to the collection direction. According to these factors, the direction deviation correction factor and the interference distribution correction factor are generated. If it is an omnidirectional collection method, it is necessary to pay attention to the noise distribution in all directions in the surrounding environment and the overall sound interference level. At the same time, it is also necessary to consider the collection distance setting parameters of the collection device, and generate the distance correction factor according to the difference between it and the optimal collection distance. In addition, according to the collection mode of the collection device (continuous collection or intermittent collection), the corresponding collection mode correction factors are generated respectively. Finally, the correction factors corresponding to these different factors are comprehensively calculated to obtain the second correction value. The second correction value fully considers the influence of the working mode and parameter setting of the sound collection device itself on noise reduction, so that the noise reduction process can be closely combined with the actual working conditions of the collection device, and the adaptability and stability of the noise reduction effect are improved.

[0069] As described in the above step S4, the first correction value and the second correction value obtained in step S3 are fused and analyzed. The two correction values ​​can be comprehensively calculated by weighted summation according to actual experience or weight coefficients determined through a large number of experiments. In the fusion process, other factors such as the overall sound complexity of the current environment, the performance parameters of the sound collection device, etc. can also be considered to dynamically adjust the weight coefficients to ensure that the fused correction value can accurately reflect the comprehensive situation of the environment and the collection device for noise reduction. In this embodiment, a fused correction value that comprehensively considers multiple factors such as noise characteristics and collection methods is obtained, which provides a basis for the subsequent accurate correction of the noise reduction algorithm parameters.

[0070] As described in the above step S5, the fusion correction value obtained in step S4 is used to correct the noise reduction algorithm parameters in the initial noise reduction mode. According to the size and direction of the fusion correction value, the various parameters of the noise reduction algorithm are adjusted accordingly, such as increasing or decreasing the intensity of noise reduction, adjusting the frequency range of noise reduction, etc. The corrected noise reduction algorithm can better adapt to the current environment and the working conditions of the acquisition device, thereby achieving more effective sound noise reduction processing. In this embodiment, the noise reduction algorithm is finally adaptively adjusted according to the actual conditions of the environment and the acquisition device, thereby improving the quality of sound acquisition and enabling people to obtain clearer and more realistic audio signals.

[0071] In this embodiment, it is possible to make dynamic adjustments in real time according to the environmental noise characteristics and the working mode of the sound collection device, adapt to various complex environments, and greatly improve the stability and reliability of the noise reduction effect. Through detailed analysis of noise characteristics and targeted correction according to the collection method, more accurate noise reduction processing can be achieved, avoiding the problem of excessive or insufficient noise reduction, and retaining the authenticity of the sound to the greatest extent. It can be applied to various types of sound collection devices, including smartphones, headphones, recording equipment, etc., and has a wide range of applicability.

[0072] In one embodiment, extracting noise feature data from the sound information includes:

[0073] Performing frequency analysis on the sound information, dividing it into different frequency ranges, and obtaining a plurality of frequency interval information;

[0074] Perform energy intensity calculation on each frequency interval information to obtain the energy intensity value of each frequency interval;

[0075] Perform noise determination processing based on the energy intensity value of each frequency interval to distinguish the noise frequency interval from the non-noise frequency interval;

[0076] The sound information in the noise frequency range is subjected to feature extraction processing to obtain noise feature data.

[0077] In this embodiment, first, a frequency analysis algorithm such as Fast Fourier Transform (FFT) is used to convert continuous sound information from the time domain to the frequency domain. According to the characteristics of human hearing and the frequency distribution characteristics of common noise, the entire audible frequency range (for example, 20Hz - 20kHz) is divided into several smaller frequency intervals. For example, it can be divided into a low-frequency interval (20Hz - 500Hz), a medium-frequency interval (500Hz - 3kHz) and a high-frequency interval (3kHz - 20kHz), etc., and each interval has a specific frequency range boundary. Decomposing complex sound information into multiple relatively simple frequency components facilitates subsequent separate analysis and processing of different frequency intervals, and can more accurately capture the distribution characteristics of noise at different frequencies.

[0078] Then, for each divided frequency interval, the energy intensity value is determined by calculating the sum of the energy of the sound signal in the interval. The energy intensity can be calculated using mathematical methods such as square sum and root mean square. The above energy intensity value reflects the activity or energy size of the sound signal in the frequency interval, and is an important indicator for determining whether the interval contains noise and the noise intensity. A higher energy intensity value means that there is greater noise interference in the interval.

[0079] Again, a series of noise determination rules and thresholds are set. For example, the frequency interval with an energy intensity value higher than a certain threshold can be preliminarily determined as an interval containing noise. At the same time, the trend of energy intensity values ​​in different frequency intervals is analyzed in combination with historical data and empirical models. If the energy intensity value of a certain frequency interval suddenly increases in a short period of time or is continuously higher than the average level, it is also listed as a suspected noise frequency interval. Then, a judgment is made based on multiple factors, such as the comparison of energy intensity with the surrounding frequency intervals, whether it conforms to the frequency characteristics of common noise, etc., to finally determine the noise frequency interval and the non-noise frequency interval. In this embodiment, the frequency interval where the noise is located is accurately separated to provide a clear target area for subsequent targeted noise reduction processing, avoid unnecessary processing of non-noise frequency intervals, and retain the integrity of useful sound signals to the greatest extent.

[0080] Finally, after determining the noise frequency interval, the sound information in these intervals is further subjected to detailed feature extraction. The features that can be extracted include but are not limited to: the duration characteristics of the noise, that is, the length of time that the statistical noise continues to appear in the frequency interval; the periodicity characteristics of the noise, judging whether the noise has a certain periodic law; the mutation characteristics of the noise, detecting whether the noise signal has a sudden change in intensity or frequency, etc. These features can be analyzed and calculated by mathematical algorithms and statistical methods. In this embodiment, the noise feature data obtained provides an accurate basis for the subsequent adjustment of the noise reduction algorithm parameters. For example, if it is found that the noise has a long duration and obvious periodicity, the parameters of the noise reduction algorithm can be adjusted in a targeted manner to better adapt to this type of noise.

[0081] In one embodiment, generating a first correction value based on the noise characteristic data includes:

[0082] Performing spectrum analysis on the noise characteristic data to determine the main frequency distribution range of the noise;

[0083] Calculating the energy intensity mean of the noise within the main frequency distribution interval;

[0084] Compare the energy intensity mean with a plurality of preset energy intensity level thresholds to determine the energy intensity level to which the noise belongs;

[0085] According to the mapping relationship between the energy intensity level and the preset correction value, a corresponding first correction value is generated.

[0086] In this embodiment, first, a spectrum analysis algorithm, such as a variant of the fast Fourier transform (FFT) algorithm or other professional audio spectrum analysis tools, is used to conduct an in-depth analysis of the noise feature data. After the noise signal is converted from the time domain to the frequency domain, the energy distribution of the noise at different frequencies can be clearly seen. By observing and analyzing the spectrum graph, the frequency areas where the energy is concentrated are found, and these areas are the main frequency distribution intervals of the noise. For example, if the energy of the noise is significantly higher than that of other frequency areas in a certain frequency range, then this frequency range is determined as the main frequency distribution interval.

[0087] After determining the main frequency distribution interval, the energy intensity of the noise signal in the interval is calculated. Common energy calculation methods can be used, such as square summing the amplitude of the noise signal sample points in the interval and then taking the average. The above energy intensity mean can quantify the overall energy level of the noise in the main frequency distribution interval and is an important indicator for measuring noise intensity. This mean will serve as the key data for determining the noise energy intensity level in the subsequent process.

[0088] A series of energy intensity level thresholds are preset, which divide the entire energy intensity range into multiple level intervals. For example, the low energy intensity level threshold can be set to T1, the medium energy intensity level threshold can be set to T2 (T2>T1), and the high energy intensity level threshold can be set to T3 (T3>T2). The calculated energy intensity mean is compared with these thresholds. If the energy intensity mean is less than T1, the noise belongs to the low energy intensity level; if it is between T1 and T2, it belongs to the medium energy intensity level; if it is greater than T2, it belongs to the high energy intensity level.

[0089] By classifying the energy intensity of noise into different levels, we can more intuitively understand the intensity level of noise, and provide a clear basis for generating corresponding correction values ​​according to different levels. Different levels of noise require different degrees of noise reduction adjustment, and the classification makes the adjustment process more accurate and reasonable.

[0090] In this embodiment, a mapping relationship table or function between energy intensity levels and correction values ​​is established. For example, for noise with low energy intensity levels, the corresponding first correction value is small, such as 0.1; for noise with medium energy intensity levels, the first correction value is moderate, such as 0.5; for noise with high energy intensity levels, the first correction value is large, such as 1.0. After the energy intensity level to which the noise belongs is determined, the corresponding first correction value is obtained by directly searching or calculating according to this mapping relationship.

[0091] The generated first correction value will be used to adjust the parameters of the noise reduction algorithm so that it can be adaptively adjusted according to the actual energy intensity of the noise. Noises of different energy intensity levels require different degrees of noise reduction. The first correction value ensures that the noise reduction algorithm can be accurately adjusted according to the intensity of the noise, improving the noise reduction effect while avoiding over-noise reduction or under-noise reduction.

[0092] In this embodiment, through the detailed analysis and classification of noise frequency and energy intensity, a first correction value that closely matches the actual characteristics of the noise can be generated, so as to achieve accurate adjustment of the noise reduction algorithm parameters. It can adapt to noises of various frequencies and intensities, whether it is slight background noise or strong interference noise, and can make reasonable parameter corrections according to its characteristics. Adjusting the noise reduction algorithm based on the precise first correction value can significantly improve the noise reduction effect, make the processed sound clearer, and reduce the interference of noise on useful sound signals.

[0093] In one embodiment, generating a first correction value based on the noise characteristic data includes:

[0094] Performing spectrum analysis on the noise characteristic data to determine a main frequency distribution range of the noise, and generating a frequency difference value as a first sub-correction value according to a difference between the main frequency distribution range and a preset standard frequency distribution range;

[0095] Analyze the noise intensity information in the noise characteristic data, match it with a plurality of preset intensity levels, determine the intensity level to which it belongs, and generate an intensity correction value as a second sub-correction value according to a correction coefficient corresponding to the intensity level;

[0096] Extracting the noise duration feature in the noise feature data, dividing it into different time periods according to a preset time segmentation rule, counting the occurrence frequency of the noise in each time period, and generating a time frequency correction value as a third sub-correction value based on the occurrence frequency;

[0097] The first correction value is obtained by performing a weighted sum calculation on the first sub-correction value, the second sub-correction value and the third sub-correction value.

[0098] In this embodiment, first, professional spectrum analysis algorithms, such as fast Fourier transform (FFT) and its improved algorithm, are used to convert noise feature data from the time domain to the frequency domain. In the frequency domain, the energy distribution of noise at different frequencies can be clearly observed. By analyzing the spectrum graph, the main frequency distribution range where the noise energy is relatively concentrated is determined.

[0099] A standard frequency distribution range suitable for most common environments is preset. This standard range can be obtained based on a large amount of experimental data and statistical analysis of common environmental noise. The determined main frequency distribution range of the noise is compared with the preset standard frequency distribution range, and the difference between the two in the frequency range is calculated. For example, a difference function can be used for calculation, such as calculating the sum or average of the energy difference between the two frequency distribution ranges at each frequency point. A numerical value is generated as the frequency difference value based on the degree of difference. The greater the difference, the greater the frequency difference value.

[0100] The first sub-correction value reflects the uniqueness of the noise frequency distribution, and can guide the noise reduction algorithm to focus on the frequency parts that are different from the standard. For example, if the main frequency distribution of the noise is in the high frequency band, and the standard frequency distribution is mainly in the medium and low frequency bands, then the larger frequency difference value will prompt the noise reduction algorithm to strengthen the noise reduction processing of the high frequency part.

[0101] Secondly, analyze the parameters related to intensity, such as amplitude and energy in the noise feature data. Compare these parameters with multiple pre-set intensity level thresholds. For example, the noise intensity can be divided into three levels: low, medium, and high, and each level corresponds to a different threshold range. When the noise intensity information meets the threshold condition of a certain level, it is determined that the noise belongs to this intensity level. For each intensity level, a corresponding correction coefficient is pre-set. These correction coefficients are derived based on a large amount of experimental and empirical data, and are used to reflect the degree of influence of different intensity levels on the noise reduction process. For example, the correction coefficient corresponding to the high intensity level is larger, which means that the noise reduction algorithm parameters need to be adjusted to a greater extent. Generate an intensity correction value based on the correction coefficient corresponding to the determined noise intensity level.

[0102] The second sub-correction value adjusts the noise reduction algorithm according to the actual intensity of the noise, ensuring that appropriate noise reduction can be performed for noises of different intensities. High-intensity noise requires stronger noise reduction, while low-intensity noise requires relatively weaker processing to avoid excessive noise reduction affecting sound quality.

[0103] Finally, the time axis of the noise feature data is divided according to the time segmentation rules. The time segmentation rules can be determined according to the specific application scenario and the characteristics of the noise. For example, the time can be divided into different time periods such as seconds and milliseconds. Then, in each time period, the number of occurrences of the noise or the proportion of the duration of the noise in the time period is counted, that is, the frequency of occurrence of the noise. For example, if the noise lasts for half of the time in a time period, the frequency of occurrence of the noise in the time period is 50%. Based on the frequency of occurrence of the noise in each time period obtained by statistics, a time frequency correction value is generated through a specific function or algorithm. For example, the frequency of occurrence can be weighted, or an adjustment value can be determined based on the distribution of the frequency of occurrence. If the noise occurs frequently in certain time periods, the time frequency correction value will be increased accordingly to prompt the noise reduction algorithm to strengthen the noise reduction processing in these time periods.

[0104] The third sub-correction value takes into account the distribution characteristics of noise in the time dimension, and enables the noise reduction algorithm to be dynamically adjusted according to the temporal variation of noise. For example, for intermittent noise, the noise reduction process can be concentrated during the period when the noise occurs, and the noise reduction intensity can be appropriately reduced during the period when the noise is not, so as to save computing resources and maintain the naturalness of the sound.

[0105] In this embodiment, a weight coefficient is assigned to each sub-correction value. These weight coefficients can be determined according to the needs of different application scenarios and the evaluation of the importance of each sub-correction value. For example, if the impact of noise intensity is relatively large in a certain scenario, the weight of the second sub-correction value can be set higher; while in another scenario, the difference in the frequency distribution of noise is more critical, then the weight of the first sub-correction value will be larger. The three sub-correction values ​​are multiplied by their respective weight coefficients and then summed to obtain the final first correction value. That is, the first correction value = first sub-correction value × weight 1 + second sub-correction value × weight 2 + third sub-correction value × weight 3. By weighted summation, the three important characteristics of noise frequency, intensity and duration are comprehensively considered to generate a first correction value that can fully reflect the noise characteristics. This correction value can more accurately guide the noise reduction algorithm to adjust parameters to achieve the best noise reduction effect.

[0106] In one embodiment, the sound collection method includes a collection direction, a collection distance setting parameter, and a collection mode.

[0107] In one embodiment, generating the second correction value based on the sound collection method includes:

[0108] If the collection direction is directional collection, a direction deviation correction factor is generated as the first part of the correction value according to the deviation between the directional angle and the preset standard angle; if it is omnidirectional collection, an interference distribution correction factor is generated as the first part of the correction value based on the position distribution of other sound interference sources in the current environment relative to the sound collection device;

[0109] Comparing the collection distance setting parameters of the sound collection device with a preset optimal collection distance range, and generating a distance correction factor as a second part correction value according to the degree of deviation;

[0110] If the acquisition mode is continuous acquisition, a continuous acquisition correction factor is generated as the third part correction value according to the length of the acquisition time and the stability of the sound data; if the acquisition mode is intermittent acquisition, an intermittent acquisition correction factor is generated as the third part correction value according to the intermittent frequency and the duration of each acquisition;

[0111] The first correction value, the second correction value and the third correction value are fused and calculated to obtain the second correction value.

[0112] In this embodiment, in the directional collection mode, the sound collection device has a specific directionality, and its collection angle is crucial for obtaining clear target sounds and reducing noise interference. First, determine the preset standard angle, which can be set according to common application scenarios and empirical values ​​of the best collection effect. Then, detect the current directional angle through built-in sensors or related technical means. Calculate the deviation value between the directional angle and the preset standard angle, and generate a directional deviation correction factor based on this deviation value. For example, a simple linear function can be used, and the larger the deviation, the larger the value of the generated correction factor. The directional deviation correction factor can reflect the degree to which the directional collection angle deviates from the optimal angle, thereby guiding the noise reduction algorithm to make targeted adjustments to the noise affected by the angle deviation. For example, if the directional angle deviation is large, more ambient noise will enter the collection range, and at this time it is necessary to enhance the ability to suppress noise in a specific direction.

[0113] If the sound collection device receives sound signals from all directions in the omnidirectional collection mode, the position distribution of other sound interference sources relative to the sound collection device has an important impact on the noise reduction effect. By using positioning technology (such as a positioning method based on signal strength or a multi-microphone array positioning technology), the position of the main sound interference source in the current environment is determined. Then, according to the relative position relationship between these interference sources and the sound collection device, the degree of their influence on the collected signal is analyzed. For example, if the interference sources are concentrated in a specific area, the noise corresponding to the area has a greater impact on the collected signal. Based on this position distribution, an interference distribution correction factor is generated. A calculation method based on position weight can be used. The closer the interference source is to the sound collection device or the greater the impact is, the higher the corresponding weight is, thereby generating a corresponding correction factor. The interference distribution correction factor enables the noise reduction algorithm to adjust the noise reduction strategy according to the actual interference source position distribution in the omnidirectional collection mode. For example, for areas with more surrounding interference sources, the noise reduction strength is enhanced, while for relatively quiet areas, the noise reduction processing is appropriately reduced to maintain the naturalness of the sound.

[0114] In this embodiment, an optimal collection distance range is also preset, and this range is determined according to the performance characteristics of the sound collection device, the law of sound propagation, and the requirements of common application scenarios. The actual collection distance between the current sound collection device and the target sound source is measured by a built-in distance sensor or other distance measurement technology. The actual collection distance is compared with the preset optimal collection distance range to calculate the degree of deviation. For example, the proportion of the difference between the actual distance and the optimal distance range to the optimal distance range can be calculated. A distance correction factor is generated according to this degree of deviation. Generally speaking, the greater the degree of deviation, the greater the value of the distance correction factor. The distance correction factor can reflect the impact of the collection distance on the quality of the sound signal and the degree of noise interference. When the collection distance deviates from the optimal range, the characteristics of the noise and the interference mode of the useful sound signal will change. The parameters of the noise reduction algorithm can be adjusted through the distance correction factor to adapt to different collection distance conditions. For example, when the collection distance is too far, it is necessary to enhance the ability to suppress the noise generated during long-distance propagation (such as noise caused by air absorption, reflection, etc.).

[0115] In this embodiment, if the sound signal is continuously recorded in the continuous acquisition mode. The cumulative effect of the length of the acquisition time on the sound signal and noise is analyzed. Long-term continuous acquisition will lead to the accumulation and change of noise, and the stability of the sound data also needs to be considered. The stability is measured by analyzing the fluctuation of the sound signal over a period of time, such as calculating the standard deviation or coefficient of variation of the sound signal. According to the length of the acquisition time and the stability of the sound data, a continuous acquisition correction factor is generated. For example, a function based on time and stability can be used. The longer the acquisition time and the worse the stability, the greater the value of the continuous acquisition correction factor. The continuous acquisition correction factor can adjust the noise reduction algorithm to adapt to the long-term acquisition requirements according to the time factor and data stability in the continuous acquisition process. For example, in the long-term acquisition process, as the noise accumulates, the noise reduction strength needs to be gradually increased. At the same time, for the period of poor stability, more sophisticated noise reduction processing is required to improve the sound quality.

[0116] If in intermittent acquisition mode, pay attention to the intermittent frequency (i.e., the time interval between acquisitions) and the duration of each acquisition. Different intermittent frequencies will affect the characteristics of noise and the interference pattern on useful sound signals. For example, high-frequency intermittent acquisition will cause the noise to change greatly between each acquisition, while low-frequency intermittent acquisition is relatively stable. The duration of each acquisition will also affect the integrity of the collected sound data and the manifestation of noise. According to the intermittent frequency and the duration of each acquisition, the intermittent acquisition correction factor is generated through a specific algorithm. For example, the intermittent frequency and the acquisition duration can be used as input parameters, and the correction factor can be calculated through a preset function. The higher the intermittent frequency and the shorter the acquisition duration, the larger the value of the correction factor. The intermittent acquisition correction factor can adjust the noise reduction algorithm to adapt to different combinations of intermittent frequencies and acquisition durations according to the characteristics of the intermittent acquisition mode. For example, for high-frequency intermittent acquisition, it is necessary to quickly adapt to the changes in noise at the beginning of each acquisition, while for longer acquisition durations, it is necessary to pay more attention to the continuous suppression of noise during the acquisition period.

[0117] Furthermore, a weight coefficient is assigned to the correction value of each part. The above weight coefficient can be determined according to the importance of each factor under different collection modes and the degree of influence on the noise reduction effect. For example, in the directional collection mode, the weight of the directional deviation correction factor is relatively high; in some specific scenarios, the influence of the collection distance or the collection mode is more critical, and the corresponding weight will also be adjusted. Then, the first part correction value, the second part correction value and the third part correction value are multiplied by their respective weight coefficients and summed to obtain the final second correction value. That is, the second correction value = the first part correction value × weight 1 + the second part correction value × weight 2 + the third part correction value × weight 3. Through fusion calculation, the influence of multiple key factors such as the collection direction, collection distance and collection mode in the sound collection mode on noise reduction is comprehensively considered to generate a second correction value that can fully reflect the characteristics of the sound collection mode. This correction value enables the noise reduction algorithm to make precise parameter adjustments according to the actual collection situation to achieve the best noise reduction effect.

[0118] In this embodiment, the noise reduction parameters can be automatically adjusted according to different sound collection methods to adapt to various complex collection environments and application scenarios, greatly improving the flexibility and practicality of the noise reduction system. Through detailed analysis of key factors in the collection method and generation of correction values, accurate suppression of noise under different collection conditions can be achieved, and sound quality and clarity can be improved. Targeted noise reduction adjustments based on the characteristics of the collection method can reasonably allocate computing resources and improve system efficiency and stability while ensuring the noise reduction effect.

[0119] In one embodiment, the initial noise reduction algorithm parameters of the noise reduction algorithm are corrected based on the fusion correction value to obtain a corrected noise reduction algorithm for performing sound noise reduction processing, including:

[0120] The sound data obtained after the noise reduction process is stored in a voice database, and query permissions are configured for the sound data;

[0121] Performing hash operations on the first correction value and the second correction value to obtain corresponding first hash values ​​and second hash values;

[0122] Add the characters in the first hash value and the second hash value to the data table of the same specification one by one in sequence to obtain the corresponding first data table and second data table;

[0123] Performing change processing on the first data table and the second data table to obtain a third data table;

[0124] Based on the third data table and a preset algorithm, a query code is generated; wherein the query code is used to verify the query authority of the sound data.

[0125] In this embodiment, a voice database is established specifically for storing sound data after noise reduction processing. The above-mentioned voice database can adopt a relational database management system (such as MySQL, Oracle, etc.) or a non-relational database (such as MongoDB, etc.), and is selected according to the characteristics of the sound data and application requirements. When the sound data is stored in the database, a unique identifier is assigned to each sound data record for subsequent retrieval and management. At the same time, different levels of query permissions are set. For example, it can be divided into administrator permissions, ordinary user permissions, etc. Administrators can view, edit and delete all sound data, while ordinary users can only view part or specific sound data. These permissions can be implemented through the user management and access control mechanism of the database, such as creating different user roles in the database and assigning corresponding permissions to each role. The orderly storage and management of the sound data after noise reduction processing is realized, and the security and privacy of the data are guaranteed through permission configuration to prevent unauthorized access and use.

[0126] Hash function is an algorithm that converts input data of any length into output data of fixed length. For the first correction value and the second correction value, use appropriate hash functions for processing respectively. Common hash functions include MD5, SHA-256, etc. The first correction value is used as input, and after calculation by the hash function, a first hash value of fixed length is obtained; similarly, the second correction value is hashed to obtain the second hash value. Due to the characteristics of the hash function, even if there is a slight change in the input data, the output hash value will be very different, and the calculation process of the hash value is irreversible, that is, the original data cannot be deduced from the hash value. Converting the first correction value and the second correction value into a hash value can, on the one hand, protect the security of the original correction value and prevent it from being directly obtained and tampered with; on the other hand, the hash value can be used as a unique identifier for subsequent data processing and permission verification processes.

[0127] In this embodiment, a data table with a preset format and structure is also created. This data table can be a two-dimensional matrix or other structure suitable for storing character data. The specifications of the data table (such as the number of rows, the number of columns, etc.) are determined according to actual needs and designs. Then, each character in the first hash value is added one by one to the corresponding position of the first data table in a certain order (for example, from left to right). For example, if the first hash value is a string composed of numbers and letters, each character can be filled in the different cells of the first data table in sequence. Similarly, the characters of the second hash value are also added to the second data table in sequence. By converting the hash value into the form of a data table, a structured method is provided for subsequent data processing and combination, which is convenient for further operation and analysis.

[0128] Further, the first data table and the second data table are changed to obtain a third data table. The change process may include multiple operations, such as merging, splicing, data conversion, etc. of the first data table and the second data table. A common method is to merge the two data tables according to specific rules. For example, the characters in the corresponding cells of the first data table and the second data table can be combined, or the rows or columns of the two data tables can be spliced ​​in a certain order. Some conversion operations can also be performed on the data in the data table, such as encoding conversion of characters, data transformation according to a specific algorithm, etc. After these change processes, a new third data table is obtained. In this embodiment, by changing the first data table and the second data table, a comprehensive data table containing information related to the first correction value and the second correction value is generated to provide a data basis for generating a query code.

[0129] Finally, based on the third data table and the preset algorithm, a query code is generated; wherein the query code is used to verify the query authority of the sound data, and the preset algorithm can be a custom encryption algorithm or a calculation method based on a specific mathematical model. The algorithm takes the data in the third data table as input, and generates a query code with a certain length and format through calculation and processing steps. For example, mathematical operations (such as sum, product, XOR, etc.) can be performed on the data in the third data table, and then combined with some random numbers or other key information, after multiple iterations and transformations, the query code is finally generated. The query code serves as a key credential for verifying the query authority of the sound data. When a user requests to access the sound data, a query code is required. The query code is verified according to the preset algorithm and the stored data. If the query code matches correctly, it means that the user has the corresponding query authority and can be allowed to access the sound data; otherwise, access is denied, thereby ensuring the security of the sound data and the legitimacy of the access.

[0130] In this embodiment, by storing the sound data in a special database and configuring the query authority, the orderly management and classification of the sound data is achieved, which is convenient for subsequent retrieval, analysis and use. By adopting hash operation and the method of generating query codes based on data tables, an innovative authority verification mechanism is provided, which improves the security of sound data access and reduces the risk of unauthorized access to data. The query authority settings and the parameters of the preset algorithm can be flexibly adjusted according to different application requirements and security levels to adapt to various complex scenarios. For some sound data involving personal privacy or sensitive information, it can be effectively protected from arbitrary access and leakage, meeting the user's demand for data privacy protection.

[0131] In one embodiment, the first data table and the second data table are changed to obtain a third data table, including:

[0132] Compare the cells at the same position of the first data table and the second data table one by one. When comparing, take the cells with the same character type at the same position of the first data table and the second data table as target cells, and delete all characters in other cells of the first data table and the second data table; if there are no cells with the same character type, take all cells as target cells;

[0133] For each target cell in the first data table, the characters in the target cell at the corresponding position in the second data table are combined with the characters in the target cell in the first data table to obtain the third data table.

[0134] In this embodiment, first, a cell-by-cell comparison mechanism is established. Since the first data table and the second data table are obtained based on the processing of the first correction value and the second correction value, they have the same specifications and structure, so they can be compared one by one according to the same row and column position. When comparing the cells at each position, the type of characters in the cells is analyzed. The character type can include numbers, letters, special symbols, etc. By identifying and judging the character type, it is determined whether the character types of the cells at the same position in the two data tables are the same. If it is found that the character types in the cells at a certain position in the first data table and the second data table are the same, then the two cells are marked as target cells. If the character types of the two cells at a certain position are different, or one of the cells is empty, then it is temporarily not marked as a target cell. If no cells with the same character type are found after comparing the entire data table, then in order to ensure the integrity of the data processing, all cells are used as target cells for subsequent processing. Through the above-mentioned detailed comparison and screening, the data units with association in the two data tables can be accurately found, providing an accurate target range for subsequent data combination, and avoiding the wrong merging of irrelevant or unrelated data.

[0135] After the target cell is determined, the character deletion operation is performed on other cells that are not marked as target cells. That is, all characters in these cells are cleared to make them empty cells. The purpose of this is to reduce the redundancy and complexity of data and only retain the information in the target cells related to subsequent data processing. By clearing irrelevant information, subsequent data processing can be made more centralized and efficient, avoiding the interference of irrelevant data on key data. Improve the simplicity and availability of data, reduce the amount of calculation and complexity of subsequent data processing, and focus data processing on target cells with potential associations.

[0136] For each target cell in the first data table that has been determined, find the target cell in the corresponding position in the second data table. Then, combine the characters in the target cell in the second data table after the characters of the target cell in the first data table in a certain order (for example, directly connected to the back). For example, if a target cell in the first data table contains the character "A", and the target cell at the corresponding position in the second data table contains the character "B", then after combining them, the position becomes "AB" in the generated third data table. By performing such a character combination operation on all target cells, a brand new third data table is finally formed. The organic fusion of the relevant information in the first data table and the second data table is realized, and a third data table containing the key information of the two data tables and having a specific association structure is generated. This third data table provides an accurate and effective data basis for subsequent applications (such as generating query codes based on it for authority verification, etc.).

[0137] In one embodiment, based on the third data table and a preset algorithm, a query code is generated, including:

[0138] Obtain the total number of blank cells in the third data table; based on the total number, generate a circle; wherein the radius of the circle^2=a+total number / 2, a is a preset value, and the column width and row height of the third data table are both one unit;

[0139] Overlaying the circle onto the third data table; wherein the center of the circle is located at the center of the third data table;

[0140] All characters in the circle are obtained and combined in sequence to obtain a combined string; the combined string is screened to obtain the query code.

[0141] In this embodiment, first, each cell in the third data table is checked and counted to determine the number of blank cells therein. Since some cells have been subjected to operations such as character combination in the previous processing, and some cells have had characters deleted or are originally empty, it is necessary to accurately count the total number of these blank cells. Then, a circle is generated according to a preset mathematical formula. A fixed preset value a is preset, and this value can be adjusted. The radius of the circle is calculated using the formula "radius of the circle^2=a+total number / 2". The radius calculation method here combines the number of blank cells with the preset value, so that the size of the circle is associated with the data state in the third data table. For example, if there are more blank cells, the radius will increase accordingly; otherwise, the radius will decrease. At the same time, considering that the column width and row height of the third data table are both one unit, this provides a standard measurement basis for the relative position and size relationship between the circle and the third data table. By converting the number of blank cells in the third data table into the radius parameter of the circle, a dynamic graphics generation mechanism based on data state is established. This circle will serve as a key area for extracting characters related to the query code later. Its size and position are closely related to the third data table, which increases the complexity and randomness of the query code generation process.

[0142] Since it is known that the column width and row height of the third data table are both one unit, the coordinates of the center position can be calculated based on the number of rows and columns and the layout. The circle generated according to the number of blank cells is superimposed in a manner in which the center of the circle coincides with the center position of the third data table. In this way, a clear relative position relationship is established between the circle and the third data table. In this process, it is necessary to ensure that the drawing and superposition of the circle are based on a unified coordinate system and unit standard in order to accurately determine the position and coverage of the circle in the third data table. The circle is accurately associated with the third data table in position, so that characters can be accurately extracted from the area covered by the circle in the future, which further enhances the close connection between the query code generation process and the third data table, and also increases the complexity and security of data processing.

[0143] Based on the superposition relationship between the circle and the third data table, the area in the third data table covered by the circle is traversed. For each cell covered by the circle, the characters therein are extracted, and these characters are read in sequence from left to right and from top to bottom. All the extracted characters are combined in the order of reading to form a continuous combined string. For example, if there are multiple cells in the area covered by the circle that contain the characters "ah", "bo", "23", etc., then the combined string obtained after sequential combination is "ahbo23". By extracting the characters in the area covered by the circle and combining them into a string, the data in the third data table is converted into a preliminary string form associated with the circle, which provides the original data basis for the subsequent generation of the query code.

[0144] Finally, a specific screening algorithm is used to process the combined string. The screening algorithm may include operations such as removing specific characters (such as common punctuation marks, spaces, etc.), intercepting specific parts of the string according to certain rules (such as taking the first or last few characters), and converting characters according to a preset encryption algorithm. For example, the spaces and some common meaningless characters in the combined string can be removed first, and then the remaining characters can be converted according to a preset encryption function, such as performing a specific mathematical operation on the ASCII value of each character and then recombining it into a new string. After these screening and processing steps, the final query code is obtained. By screening and processing the combined string, it is converted into a string that meets the query code requirements and has a certain degree of security and uniqueness. This query code not only contains the information in the third data table (extracted by the association between the circle and the third data table), but also undergoes a complex processing process, making it difficult to be easily guessed or cracked, and has high security and reliability for subsequent access verification and other operations.

[0145] In this embodiment, by combining the structure and data state of the third data table with the generation of the circle and the extraction of characters, the generated query code has high randomness and complexity, which greatly increases the difficulty of cracking and improves the security of data access. The generation of the query code is directly based on the third data table, so that the query code and the original data have an intrinsic connection, ensuring that only operations related to the specific third data table can correctly generate and use the query code, thereby enhancing the security and manageability of the data.

[0146] Reference Figure 2 In another embodiment of the present invention, a system for adaptively adjusting a noise reduction mode is provided, comprising:

[0147] An extraction unit, configured to extract noise characteristic data from the sound information collected in the environment by the sound collection device;

[0148] An acquisition unit, configured to acquire an initial noise reduction mode configured by the sound collection device; wherein the initial noise reduction mode includes a noise reduction algorithm and its corresponding initial noise reduction algorithm parameters;

[0149] A generating unit, configured to generate a first correction value based on the noise characteristic data; obtain a sound collection method of the sound collection device, and generate a second correction value based on the sound collection method;

[0150] a fusion unit, configured to perform fusion analysis based on the first correction value and the second correction value to obtain a fusion correction value;

[0151] The noise reduction unit is used to modify the initial noise reduction algorithm parameters of the noise reduction algorithm based on the fusion correction value to obtain a modified noise reduction algorithm for sound noise reduction processing.

[0152] In this embodiment, for the specific implementation of each unit in the above system embodiment, please refer to the above method embodiment, which will not be repeated here.

[0153] The present invention also provides a computer chip, which is used to implement the steps of any of the above methods when executing a computer program.

[0154] In summary, the method, system and chip for adaptively adjusting the noise reduction mode provided in the embodiments of the present invention include: extracting noise feature data from the sound information in the sound collection environment based on the sound collection device; obtaining the initial noise reduction mode configured by the sound collection device; wherein the initial noise reduction mode includes a noise reduction algorithm and its corresponding initial noise reduction algorithm parameters; generating a first correction value based on the noise feature data; obtaining the sound collection method of the sound collection device, and generating a second correction value based on the sound collection method; performing fusion analysis based on the first correction value and the second correction value to obtain a fusion correction value; correcting the initial noise reduction algorithm parameters of the noise reduction algorithm based on the fusion correction value to obtain a corrected noise reduction algorithm for sound noise reduction processing. In the present invention, based on the characteristic data of the noise and the collection method of the sound collection device, the noise reduction parameters of the noise reduction mode are adaptively adjusted to achieve better noise reduction effect, thereby overcoming the defect that the current noise reduction method is unreasonable and cannot effectively remove background noise interference.

[0155] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media provided by the present invention and used in the embodiments may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double-speed data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM.

[0156] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, device, article or method. In the absence of further restrictions, an element defined by the sentence "includes a ..." does not exclude the presence of other identical elements in the process, device, article or method including the element.

[0157] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for adaptively adjusting a noise reduction mode, characterized in that: The following steps are involved: The sound collection device collects sound information in the environment, and extracts noise characteristic data from the sound information; Acquire an initial noise reduction mode configured by the sound collection device; wherein the initial noise reduction mode includes a noise reduction algorithm and its corresponding initial noise reduction algorithm parameters; Based on the noise characteristic data, a first correction value is generated; a sound collection method of the sound collection device is obtained, and a second correction value is generated based on the sound collection method; Perform fusion analysis based on the first correction value and the second correction value to obtain a fusion correction value; The initial noise reduction algorithm parameters of the noise reduction algorithm are corrected based on the fusion correction value to obtain a corrected noise reduction algorithm for sound noise reduction processing.

2. The method for adaptively adjusting the noise reduction mode according to claim 1, characterized in that: The extracting noise characteristic data from the sound information comprises: Performing frequency analysis on the sound information, dividing it into different frequency ranges, and obtaining a plurality of frequency interval information; Perform energy intensity calculation on each frequency interval information to obtain the energy intensity value of each frequency interval; Perform noise determination processing based on the energy intensity value of each frequency interval to distinguish the noise frequency interval from the non-noise frequency interval; The sound information in the noise frequency range is subjected to feature extraction processing to obtain noise feature data.

3. The method for adaptively adjusting the noise reduction mode according to claim 1, characterized in that: The step of generating a first correction value based on the noise characteristic data comprises: Performing spectrum analysis on the noise characteristic data to determine the main frequency distribution range of the noise; Calculating the energy intensity mean of the noise within the main frequency distribution interval; Compare the energy intensity mean with a plurality of preset energy intensity level thresholds to determine the energy intensity level to which the noise belongs; According to the mapping relationship between the energy intensity level and the preset correction value, a corresponding first correction value is generated.

4. The method for adaptively adjusting the noise reduction mode according to claim 1, characterized in that: The step of generating a first correction value based on the noise characteristic data comprises: Performing spectrum analysis on the noise characteristic data to determine a main frequency distribution range of the noise, and generating a frequency difference value as a first sub-correction value according to a difference between the main frequency distribution range and a preset standard frequency distribution range; Analyze the noise intensity information in the noise characteristic data, match it with a plurality of preset intensity levels, determine the intensity level to which it belongs, and generate an intensity correction value as a second sub-correction value according to a correction coefficient corresponding to the intensity level; Extracting the noise duration feature in the noise feature data, dividing it into different time periods according to a preset time segmentation rule, counting the occurrence frequency of the noise in each time period, and generating a time frequency correction value as a third sub-correction value based on the occurrence frequency; The first correction value is obtained by performing a weighted sum calculation on the first sub-correction value, the second sub-correction value and the third sub-correction value.

5. The method for adaptively adjusting the noise reduction mode according to claim 1, characterized in that: The sound collection method includes collection direction, collection distance setting parameters, and collection mode.

6. The method for adaptively adjusting the noise reduction mode according to claim 5, characterized in that: The generating a second correction value based on the sound collection method includes: If the collection direction is directional collection, a direction deviation correction factor is generated as the first part of the correction value according to the deviation between the directional angle and the preset standard angle; if it is omnidirectional collection, an interference distribution correction factor is generated as the first part of the correction value based on the position distribution of other sound interference sources in the current environment relative to the sound collection device; Comparing the collection distance setting parameters of the sound collection device with a preset optimal collection distance range, and generating a distance correction factor as a second part correction value according to the degree of deviation; If the acquisition mode is continuous acquisition, a continuous acquisition correction factor is generated as the third part correction value according to the length of the acquisition time and the stability of the sound data; if the acquisition mode is intermittent acquisition, an intermittent acquisition correction factor is generated as the third part correction value according to the intermittent frequency and the duration of each acquisition; The first correction value, the second correction value and the third correction value are fused and calculated to obtain the second correction value.

7. The method for adaptively adjusting the noise reduction mode according to claim 1, characterized in that: The initial noise reduction algorithm parameters of the noise reduction algorithm are corrected based on the fusion correction value to obtain a corrected noise reduction algorithm for performing sound noise reduction processing, comprising: The sound data obtained after the noise reduction process is stored in a voice database, and query permissions are configured for the sound data; Performing hash operations on the first correction value and the second correction value respectively to obtain corresponding first hash values ​​and second hash values; Add the characters in the first hash value and the second hash value to the data table of the same specification one by one in sequence to obtain the corresponding first data table and second data table; Performing change processing on the first data table and the second data table to obtain a third data table; Based on the third data table and a preset algorithm, a query code is generated; wherein the query code is used to verify the query authority of the sound data.

8. A system for adaptively adjusting noise reduction mode, characterized in that: include: An extraction unit, configured to extract noise characteristic data from the sound information collected in the environment by the sound collection device; An acquisition unit, configured to acquire an initial noise reduction mode configured by the sound collection device; wherein the initial noise reduction mode includes a noise reduction algorithm and its corresponding initial noise reduction algorithm parameters; A generating unit, configured to generate a first correction value based on the noise characteristic data; obtain a sound collection method of the sound collection device, and generate a second correction value based on the sound collection method; a fusion unit, configured to perform fusion analysis based on the first correction value and the second correction value to obtain a fusion correction value; The noise reduction unit is used to modify the initial noise reduction algorithm parameters of the noise reduction algorithm based on the fusion correction value to obtain a modified noise reduction algorithm for sound noise reduction processing.

9. A computer chip, characterized in that: The computer chip is used to implement the steps of the method according to any one of claims 1 to 7 when executing a computer program.

Images