Audio product noise detection method, device, equipment and storage medium

Abstract

The application discloses an audio product noise detection method and device, equipment and a storage medium. The method comprises the following steps: receiving a sound signal of an audio product to be detected, and converting the sound signal into a digital signal; performing time domain feature conversion on the digital signal through an acoustic model to obtain a characteristic curve; comparing each characteristic curve with a preset target curve to determine a target frequency band; setting a lower threshold in the target frequency band, and determining that the audio product to be detected is qualified in the case that each characteristic curve meets the condition of being above the lower threshold.

Description

Technical Field

[0001] This application relates to the field of communication technology, and in particular to methods, apparatus, devices and storage media for detecting noise in audio products. Background Technology

[0002] With the rapid development of consumer electronics, users' demands for audio product sound quality are constantly increasing, prompting manufacturers to conduct rigorous audio testing during the finished product assembly process. Modern audio products not only need to meet basic specifications such as frequency response, distortion, and higher harmonics, but also require effective detection of various acoustic defects such as abnormal noise, vibration, and air leakage. The detection of these defects relies on high-precision acoustic acquisition equipment and professional analysis algorithms. With the expansion of production scale and the diversification of product models, the demand for automation and standardization in audio testing is becoming increasingly urgent. However, existing production lines mainly rely on manual listening to detect abnormal noise. This method is affected by environmental conditions and human factors, resulting in high rates of missed detections, high rates of false detections, and high personnel training costs. While automatic listening equipment can improve detection efficiency, the equipment cost is as high as 100,000 yuan, nearly twice that of ordinary acoustic equipment, and it is difficult to use and maintain, requiring the adjustment and setting of test parameters for different product types, speaker sizes, and speaker materials.

[0003] Therefore, how to reduce the false negative and false positive rates of noise detection in audio products and reduce reliance on manual methods is an urgent problem that needs to be solved.

[0004] The above content is only used to help understand the technical solution of this application and does not represent an admission that the above content is prior art. Summary of the Invention

[0005] The main purpose of this application is to provide a method, apparatus, device, and storage medium for detecting noise in audio products, aiming to solve the technical problem of how to reduce the false negative rate and false positive rate of noise detection in audio products.

[0006] To achieve the above objectives, this application proposes a method for detecting noise in audio products, the method comprising: Receive the sound signal from the audio product under test and convert the sound signal into a digital signal; The digital signal is subjected to time-domain feature transformation using an acoustic model to obtain the feature curve; Each of the aforementioned characteristic curves is compared with a preset target curve to determine the target frequency band; A lower limit threshold is set for the target frequency band. If all the characteristic curves are above the lower limit threshold, the audio product under test is deemed qualified.

[0007] Furthermore, to achieve the above objectives, this application also proposes an audio product noise detection device, which includes: The acquisition module is used to receive the sound signal of the audio product under test and convert the sound signal into a digital signal; The feature extraction module is used to perform time-domain feature transformation on the digital signal using an acoustic model to obtain feature curves; The comparison module is used to compare each of the feature curves with a preset target curve to obtain the comparison result; The determination module is used to set a lower limit in the target frequency band based on the comparison results; if all the characteristic curves are above the lower limit, the audio product under test is determined to be qualified.

[0008] In addition, to achieve the above objectives, this application also proposes an audio product noise detection device, the device including: a memory, a processor, and an audio product noise detection program stored in the memory and executable on the processor, the audio product noise detection program being configured to implement the steps of the audio product noise detection method described above.

[0009] In addition, to achieve the above objectives, this application also proposes a storage medium, which is a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the steps of the audio product noise detection method described above.

[0010] In addition, to achieve the above objectives, this application also provides a computer program product, which includes a computer program that, when executed by a processor, implements the steps of the audio product noise detection method described above.

[0011] This application provides a method for detecting noise in audio products. The method involves receiving the sound signal of the audio product under test and converting it into a digital signal; performing time-domain feature transformation on the digital signal using an acoustic model to obtain a feature curve; comparing each feature curve with a preset target curve to determine the target frequency band; setting a lower limit threshold in the target frequency band; and determining that the audio product under test is qualified if all feature curves are above the lower limit threshold.

[0012] In summary, this application automatically performs time-domain feature transformation and multi-curve comparison analysis through acoustic models, which significantly reduces the false negative and false positive rates of manual listening detection. It can achieve standardized testing without professional personnel, and supports multi-channel parallel testing, thereby improving production line testing efficiency, effectively preventing defective products from entering the market, and greatly reducing equipment costs and personnel training costs. Attached Figure Description

[0013] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0014] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0015] Figure 1 This is a flowchart illustrating an embodiment of the audio product noise detection method of this application. Figure 2 A simplified block diagram of single-packet data verification provided for an embodiment of the audio product noise detection method of this application; Figure 3 This application provides an external Flash programming design flowchart for an embodiment of the audio product noise detection method. Figure 4 This is a flowchart illustrating Embodiment 2 of the audio product noise detection method of this application; Figure 5 This is a schematic diagram of the module structure of the audio product noise detection device according to an embodiment of this application; Figure 6 This is a schematic diagram of the device structure of the hardware operating environment involved in the audio product noise detection method of this application embodiment.

[0016] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0017] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.

[0018] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0019] The main solution of this application embodiment is: receiving the sound signal of the audio product under test and converting the sound signal into a digital signal; performing time-domain feature transformation on the digital signal through an acoustic model to obtain a feature curve; comparing each feature curve with a preset target curve to determine the target frequency band; setting a lower limit threshold in the target frequency band, and determining that the audio product under test is qualified when all feature curves are above the lower limit threshold.

[0020] With the rapid development of consumer electronics, users' demands for audio product sound quality are constantly increasing, prompting manufacturers to conduct rigorous audio testing during the finished product assembly process. Modern audio products not only need to meet basic specifications such as frequency response, distortion, and higher harmonics, but also require effective detection of various acoustic defects such as abnormal noise, vibration, and air leakage. The detection of these defects relies on high-precision acoustic acquisition equipment and professional analysis algorithms. With the expansion of production scale and the diversification of product models, the demand for automation and standardization in audio testing is becoming increasingly urgent. However, existing production lines mainly rely on manual listening to detect abnormal noise. This method is affected by environmental conditions and human factors, resulting in high rates of missed detections, high rates of false detections, and high personnel training costs. While automatic listening equipment can improve detection efficiency, the equipment cost is as high as 100,000 yuan, nearly twice that of ordinary acoustic equipment, and it is difficult to use and maintain, requiring the adjustment and setting of test parameters for different product types, speaker sizes, and speaker materials.

[0021] Therefore, how to reduce the false negative and false positive rates of noise detection in audio products and reduce reliance on manual methods is an urgent problem that needs to be solved.

[0022] This application receives the sound signal from the audio product under test and converts it into a digital signal; it performs time-domain feature transformation on the digital signal using an acoustic model to obtain characteristic curves; it compares each characteristic curve with a preset target curve to determine the target frequency band; it sets a lower threshold in the target frequency band, and if all characteristic curves are above the lower threshold, the audio product under test is deemed qualified. By automatically performing time-domain feature transformation and multi-curve comparison analysis through the acoustic model, the application significantly reduces the false negative and false positive rates of manual listening inspection, enables standardized testing without the need for professional personnel, and supports multi-channel parallel testing, improving production line inspection efficiency, effectively preventing defective products from entering the market, and significantly reducing equipment and personnel training costs.

[0023] Based on this, embodiments of this application provide a method for detecting noise in audio products, referring to... Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the audio product noise detection method of this application.

[0024] In this embodiment, the noise detection method for audio products includes steps S10 to S40: Step S10: Receive the audio signal from the audio product under test and convert the audio signal into a digital signal; like Figure 2 As shown, Figure 2 For demonstration purposes, this solution is applied to an acoustic curve detection system for abnormal noise, which includes a sound card acquisition device, a standard microphone sensor, a Bluetooth adapter, a product placement location, a power amplifier, an acoustically isolated soundproof enclosure, and a host computer.

[0025] like Figure 3 As shown, Figure 3 The test flowchart includes: opening the soundproof box, placing the audio device under test to start the test, acquiring, playing, analyzing and processing data through the sound card to obtain the test results, retesting if the test result is unqualified, and retesting a second time if the retest is still unqualified. When the test is passed, the soundproof box is opened and the process ends.

[0026] It should be noted that the audio product under test refers to an audio playback device that requires noise detection; in this embodiment, it specifically refers to consumer electronics products such as headphones, soundbars, and wireless audio products. The sound signal refers to the continuous analog sound wave vibrations generated by the audio product under test during operation, containing all acoustic information when the product plays audio. The digital signal refers to the binary data stream obtained by discretizing the continuous analog sound signal using analog-to-digital conversion technology, facilitating subsequent computer processing and analysis.

[0027] As is understandable, this step first involves placing the audio product under test into a soundproof enclosure. Once the enclosure detects the product and places it in position, it sends a trigger signal to the host computer. Upon receiving this signal, the host computer activates the sound card to begin data acquisition. Subsequently, the standard microphone sensor picks up the raw analog sound signal generated by the audio product during operation. The standard microphone sensor transmits the acquired analog sound signal back to the sound card, which converts the analog signal into a digital sound signal using an analog-to-digital converter. Finally, the converted digital sound signal is transmitted to the host computer for storage and further processing. This step achieves a complete conversion process from physical sound to digital data, providing a fundamental data source for subsequent acoustic analysis. This ensures that subsequent analysis is based on accurate and reliable digital signals, avoiding attenuation and interference issues that occur during analog signal transmission.

[0028] In practice, the soundproof box is opened and the product is placed inside. After receiving the signal, the host computer starts the sound card to collect the audio signal. The microphone sensor begins to pick up the sound signal generated during the operation of the audio product. The sound card converts the analog signal into a digital signal and sends it back to the host computer on the PC for analysis, processing and judgment.

[0029] In one feasible implementation, step S10 may include steps A11-A12: Step A11: The soundproof box detects and forwards the acquisition signal emitted by the audio product under test to the host computer to control the microphone sensor to pick up the sound signal generated by the audio product under test during operation. It should be noted that the soundproof enclosure, specifically the high-isolation acoustic enclosure, provides a testing environment with high isolation from external noise in this embodiment, ensuring that the acquired sound signals are not interfered with by production line noise. The standard microphone sensor is a standardized microphone sensor with stable and sensitive acoustic pickup capabilities, used to accurately acquire the weak acoustic signals emitted by the product under test. The host computer refers to the computer device running the acoustic analysis software, responsible for controlling the entire testing process and data processing.

[0030] As is understandable, this step begins by opening the soundproof box and placing the audio product under test inside. Once the soundproof box detects that the product is in place, it sends a signal to the host computer. Upon receiving this signal, the host computer immediately activates the sound card to begin data acquisition. Subsequently, the microphone sensor picks up the sound signals generated by the audio product during operation. The microphone sensor transmits the acquired analog sound signals back to the sound card, which then converts the analog signal into a digital sound signal that can be processed by the computer. Finally, the digital sound signal is transmitted back to the host computer for storage. This step achieves a complete conversion process from physical sound to digital data, providing a fundamental data source for subsequent acoustic analysis. This ensures that subsequent analysis is based on accurate and reliable digital signals, avoiding attenuation and interference issues during analog signal transmission.

[0031] Step A12: The audio signal is transmitted to the sound card to convert the audio signal into a digital audio signal.

[0032] Understandably, this step transmits the analog sound signal picked up by the microphone sensor to the sound card, and controls the sound card to convert the analog signal into a digital signal through the analog-to-digital conversion function. The converted digital sound signal can be recognized and processed by the host computer, providing digital input for subsequent acoustic model analysis.

[0033] Step S20: Perform time-domain feature transformation on the digital signal using an acoustic model to obtain the feature curve; It should be noted that the acoustic model refers to a mathematical algorithm model established through training on a large number of samples. In this embodiment, multiple signal processing algorithms are integrated for automated analysis and feature extraction of audio signals. Time-domain feature transformation refers to the process of transforming a continuous signal in the time domain to other analysis domains, namely the frequency domain or the Bark domain, through mathematical transformation, in order to more effectively identify abnormal features in the signal. Feature curves refer to multi-dimensional acoustic feature curves extracted from the original signal. In this embodiment, these specifically include intensity spectrum curves, frequency spectrum curves, loudness curves, and sharpness curves. These curves together constitute the acoustic fingerprint of the product.

[0034] Understandably, this step first inputs the digital signal into the acoustic model. The acoustic model performs normalized intensity spectrum analysis on the digital signal, obtaining an intensity spectrum curve reflecting the signal intensity distribution through short-time Fourier transform and energy calculation. Simultaneously, a fast Fourier transform is performed to obtain a spectrum curve, which shows the energy distribution of the signal at different frequencies. Then, characteristic loudness and sharpness analysis are performed, calculating the loudness curve perceived by the human ear and the sharpness curve reflecting the stimuliness of high-frequency components based on the Zwicker psychoacoustic model. Finally, these curves are combined as the feature curve output. This step, through multi-dimensional feature extraction, transforms the original sound signal into feature curves with clear physical and auditory meaning, providing rich feature information for subsequent accurate comparison and identification, and significantly improving the detection capability of abnormal noise.

[0035] In the specific implementation, based on the characteristic signals of several good and defective products initially screened and distinguished by the acoustic model, a second time-domain feature transformation is performed: The system utilizes a self-developed normalized intensity spectrum analysis method to better distinguish differences. The signal is segmented, and to reduce spectral leakage, each segment requires windowing. This system employs a Hamming window approach, using short-time Fourier transform calculations to square the real part and calculate the energy at different frequencies for each segment. The results are then processed using the mean value and output. Within a specified frequency band, the intensity spectrum from 0 to 24 bars is calculated through nonlinear normalization.

[0036] The time-domain waveform is converted into the frequency domain and Bark domain for analysis and comparison by using FFT spectrum analysis, loudness and sharpness analysis.

[0037] Before step S20, the following are also included: In one feasible implementation, step S20 may include steps A21 to A26: Step A21: Obtain the original sound signals generated during the operation of a preset number of audio products; In the specific implementation, the acoustic model construction includes: after preliminary judgment of the noise test sound, normal sound and abnormal sound are close to the mid-high frequency range, so bandpass filtering can be used to initially filter out irrelevant signals (200-10000 bandpass filtering) to make subsequent analysis more effective and accurate.

[0038] By using custom feature extraction and calibration methods, and employing autocorrelation and cross-correlation comparison algorithms, as well as signal similarity methods, several types of undesirable sound feature signals are initially screened and compared with normal sound signals to establish a corresponding feature sample library.

[0039] The model parameter tuning process includes: in the early stage, collecting a large amount of data to expand the training model samples, comparing the similarity and correlation values ​​between each signal and the good product signal, continuously optimizing and adjusting the parameters, selecting the optimal signal feature template, and improving the accuracy of model recognition.

[0040] It should be noted that the raw sound signal refers to the analog sound waveform directly collected from the product under test without any processing.

[0041] Understandably, this step involves acquiring a preset number of raw sound signals generated during the operation of audio products. These signals include known samples of good and defective products, serving as the data basis for training the acoustic model.

[0042] Step A22: Convert the analog signal of the original sound signal into a digital signal for training using the sound card; Understandably, this step uses a sound card to convert the original analog audio signal into a digital signal for training, ensuring a consistent data format for easier subsequent digital processing.

[0043] Step A23: Perform bandpass filtering on the training digital signal to retain the target training signal within the preset frequency range; It should be noted that bandpass filtering is a signal processing technique that allows signals within a specific frequency range to pass through while filtering out other frequency components. In this embodiment, it is used to retain effective signals in the mid-to-high frequency band.

[0044] Understandably, this step involves bandpass filtering of the training digital signal. In this embodiment, the target training signal within a preset frequency range is retained, while irrelevant low-frequency and high-frequency interference components are filtered out, making subsequent analysis more effective and accurate.

[0045] Step A24: Perform custom feature extraction and calibration on the target training signal to obtain the sound feature parameters of the target training signal; Understandably, this step involves custom feature extraction and calibration of the target training signal. Sound feature parameters are extracted using normalized intensity spectrum analysis, fast Fourier transform, and characteristic loudness and sharpness analysis methods to describe the signal's spectral characteristics, temporal characteristics, and auditory characteristics.

[0046] Step A25: Through autocorrelation comparison algorithm, cross-correlation comparison algorithm and signal similarity analysis, the sound feature parameters are screened and compared to separate the good product feature signals and the defective product feature signals. Understandably, this step uses autocorrelation comparison algorithm (comparing the similarity of a signal to itself), cross-correlation comparison algorithm (comparing the similarity between different signals), and signal similarity analysis to screen and compare sound feature parameters, thereby separating good product feature signals from defective product feature signals and establishing a corresponding feature sample library.

[0047] Step A26: Store the good product feature signals and defective product feature signals into the feature sample library, expand the training sample size, compare the similarity and correlation between each training sample and the good product feature signals, iteratively optimize the model parameters, determine the target signal feature template, and obtain the trained acoustic model.

[0048] It should be noted that the feature sample library refers to a dataset storing feature signals of good and defective products, used for model training and parameter optimization. Iterative optimization refers to the process of repeatedly adjusting model parameters to improve performance. The target signal feature template refers to the optimal feature pattern determined through training to distinguish between good and defective products.

[0049] Understandably, this step stores the separated good and defective product feature signals into a feature sample library to expand the training sample size. By comparing the similarity and correlation between each training sample and the good product feature signal, the parameters of the acoustic model are iteratively optimized, and finally the optimal target signal feature template is determined, resulting in the trained acoustic model. This training process establishes the ability to distinguish between good and defective products through learning from a large number of samples, providing a reliable model foundation for online detection and significantly improving the accuracy and stability of subsequent detection.

[0050] Step S30: Compare each characteristic curve with the preset target curve to determine the target frequency band; It should be noted that the preset gold sample refers to a benchmark qualified product that has been verified and confirmed to be free of noise through both manual and instrumental methods, and its acoustic characteristics serve as the judgment criterion. The benchmark characteristic curve refers to the set of characteristic curves of various dimensions extracted from the gold sample, including the benchmark intensity spectrum curve, benchmark frequency spectrum curve, benchmark loudness curve, and benchmark sharpness curve. Similarity data refers to numerical values ​​that quantify the degree of similarity between the characteristics of the product to be tested and the characteristics of the gold sample.

[0051] It can be understood that in this step, the reference feature curves of the preset gold sample are first obtained. The reference feature curves include the reference intensity spectrum curve, the reference frequency spectrum curve, the reference loudness curve, and the reference sharpness curve. These curves constitute the standard acoustic fingerprint of the product. Subsequently, the similarity data between each feature curve and the reference feature curve in each corresponding frequency band is calculated, that is, the correlation value is calculated through the autocorrelation algorithm and the cross-correlation algorithm, and the similarity value is obtained through the signal similarity method. Finally, the similarity data is compared with the preset similarity threshold, and the target frequency band, that is, the frequency band where the defective product and the non-defective product have a large difference, is screened out. Through the multi-dimensional curve comparison and similarity calculation in this comparison process, the frequency band with significant differences can be accurately located, providing a scientific basis for setting the determination lower limit in the subsequent process, and ensuring the accuracy and pertinence of the detection.

[0052] In a feasible implementation manner, step S30 may include steps A31 to A33: Step A31, obtain the reference feature curves of the preset gold sample, where the reference feature curves include the reference intensity spectrum curve, the reference frequency spectrum curve, the reference loudness curve, and the reference sharpness curve; It can be understood that in this step, the reference feature curves of the preset gold sample are obtained. The reference feature curves include the reference intensity spectrum curve, the reference frequency spectrum curve, the reference loudness curve, and the reference sharpness curve. These curves constitute the standard acoustic fingerprint of the product and serve as the reference for subsequent comparison.

[0053] Step A32, calculate the similarity data between each feature curve and the reference feature curve in each corresponding frequency band; It can be understood that in this step, the similarity data between each feature curve and the reference feature curve in each corresponding frequency band is calculated, that is, the correlation value is calculated through the autocorrelation algorithm and the cross-correlation algorithm, and the similarity value is obtained through the signal similarity method, which is used to quantify the difference degree between the待测产品 (to-be-tested product) and the gold sample.

[0054] Step A33, compare the similarity data with the preset similarity threshold, and screen out the target frequency band; It can be understood that in this step, the similarity data is compared with the preset similarity threshold, and the target frequency band, that is, the frequency band where the defective product and the non-defective product have a large difference, is screened out, providing a basis for setting the determination lower limit in the subsequent process.

[0055] Step S40, set the lower limit threshold in the target frequency band. If each feature curve satisfies being above the lower limit threshold, it is determined that the to-be-tested audio product is qualified.

[0056] It should be noted that the lower limit threshold refers to the lowest limit value for determining the qualification of the product, and it is set in the frequency band with a large difference to distinguish between non-defective and defective products. The determination rule refers to the product quality determination criterion that includes the lower limit threshold and its application logic. The verification by frequency band refers to the process of checking and confirming one by one for each frequency band.

[0057] Understandably, this step first sets a lower threshold within the target frequency band, i.e., the selected frequency bands with large differences, thus obtaining a clear judgment rule. Then, each characteristic curve is verified band by band according to this judgment rule, that is, the value of each characteristic curve in that frequency band is checked to see if it meets the lower threshold requirement. When all characteristic curves are above the lower threshold, the audio product under test is judged to be qualified. This judgment process, through setting a dynamic lower threshold and a strict band-by-band verification mechanism, ensures the scientific rigor and precision of the judgment standard, effectively avoiding missed detections and false judgments, and improving the reliability of the test.

[0058] In the specific implementation, multiple sets of curves are obtained through the analysis of the above algorithms. By comparing the curve results, it can be seen that there is a large difference between defective products and good products in a certain frequency band / Bark band. Therefore, a lower limit is set in this frequency band. All curves are judged to be qualified if they are above the lower limit, otherwise they are judged to be unqualified.

[0059] In one feasible implementation, step S40 may include steps A34-A36: Step A34: Set a lower limit threshold for the target frequency band to obtain the judgment rule; Understandably, this step sets a lower threshold for the target frequency band to obtain the judgment rule. This rule clarifies the qualification standard for each frequency band and provides a basis for product judgment.

[0060] Step A35: Verify each characteristic curve band by band according to the judgment rules. When each characteristic curve is above the lower limit threshold, the audio product under test is judged to be qualified. Understandably, this step verifies each characteristic curve band by band according to the judgment rules. When each characteristic curve is above the lower limit threshold, the audio product under test is deemed qualified, ensuring that the product meets the quality standards.

[0061] Step A36: When all characteristic curves are below the lower threshold, the audio product under test is determined to be unqualified, and the process returns to the step of performing time-domain feature transformation on the digital signal through the acoustic model to obtain the characteristic curves.

[0062] Understandably, this step determines that the audio product under test is unqualified when all characteristic curves are below the lower threshold, and returns to the step of performing time-domain feature transformation on the digital signal through the acoustic model for retesting, thus improving the reliability of the detection.

[0063] It should be understood that the test system is an automated test software customized by Labview. It can support the testing of various audio products or other semi-finished products. Through the simple drag-and-drop operation of the sequence, the test method and test items can be arbitrarily selected, the test steps can be added or deleted at will, the analysis algorithm can be freely selected, and the test results are presented in the form of charts, including the records of all test data, the saving of Logs, the saving of pictures, etc., which is convenient for later traceability and data viewing. It supports its own MES system and can synchronously upload data and results to the MES server.

[0064] The measurement system of this embodiment includes hardware components such as a high-isolation box, a data acquisition card (i.e., a MORY sound card device), an audio transmission line, a signal amplifier, and a low-noise artificial ear. The hardware configuration supports the setting of communication interaction ports and the selection of sound card types, including parameter adjustments such as channel settings, sampling rates, and sensitivity settings. The test system integrates a variety of analysis algorithms, can select the best analysis method according to specific scenarios and specific states, and the algorithm analysis supports the simultaneous analysis of multiple types. It can freely combine analysis methods such as fast Fourier transform (FFT), short-time Fourier transform (STFT), energy spectrum, intensity spectrum, sound pressure level, loudness, and sharpness. The threshold setting function allows the selection of a reference gold sample as the target curve and saves it in the local data. The actual test prototype is based on this. If the test curve is within the range of the gold machine target curve, it is considered qualified, and if it exceeds, it represents a test failure. The data saving function supports the real-time one-key saving of all test curves, test Logs, test time-domain waveforms, and interface screenshots of the current test item. The local gold machine Logs are saved separately for easy subsequent search and traceability.

[0065] This embodiment provides a method for detecting noise in audio products. By automatically performing time-domain feature conversion and multi-curve comparison analysis through an acoustic model, the missed detection rate and false detection rate of manual listening detection are significantly reduced, and the personnel training cost and equipment investment cost are greatly reduced. It supports multi-channel parallel testing to synchronously detect multiple products, effectively improving the detection efficiency of the production line. By presetting the gold sample target curve and judgment rules, the unification and objectification of detection standards are achieved, avoiding the false detection risk caused by environmental conditions and human factors in manual detection. This embodiment also supports uploading test data and results to the manufacturing execution system server, which is convenient for later traceability and quality analysis, effectively preventing defective products from flowing into the market, and improving product yield and customer satisfaction.

[0066] Based on the first embodiment of this application, in the second embodiment of this application, the same or similar content as the above-mentioned embodiment one can be referred to the above introduction and will not be repeated hereinafter. On this basis, please refer to Figure 4 , step S20 further includes steps S201 to S204: Step S201, input the digital signal into the acoustic model, perform normalized intensity spectrum analysis on the digital signal, and obtain the intensity spectrum curve; It should be noted that the acoustic model refers to a mathematical algorithm model established through training on a large number of samples. In this embodiment, multiple signal processing algorithms are integrated for automated analysis and feature extraction of audio signals. Normalized intensity spectrum analysis is a method of extracting spectral features after standardizing the signal intensity, used to eliminate the influence of signal amplitude differences on the analysis results. The intensity spectrum curve is a continuous curve reflecting the intensity distribution of the signal in different frequency bands, and is an important basis for subsequent comparison and judgment.

[0067] Understandably, this step involves inputting the digital signal into the acoustic model. The acoustic model segments the digital signal, applies a window function to each segment, performs a short-time Fourier transform, calculates the energy spectrum, and performs nonlinear normalization, ultimately generating the intensity spectrum curve in the Bark domain. This process converts the time-domain signal into a frequency-domain representation using time-frequency analysis techniques, effectively extracting the energy distribution characteristics of the signal in different frequency bands, providing a crucial data foundation for subsequent anomaly detection.

[0068] In one feasible implementation, step S201 may include steps A41 to A45: Step A41: In the acoustic model, the digital signal is segmented and processed, and a window function is applied to each signal segment to obtain a windowed signal segment; It should be noted that a window function is a mathematical function used in signal processing to weight signals. In this embodiment, a Hamming window is used to reduce spectral leakage. The windowed signal segment refers to the signal segment after being weighted by the window function, preparing it for subsequent Fourier transform.

[0069] Understandably, this step divides the digital signal into multiple continuous signal segments in the acoustic model, applies a window function to each segment, and obtains a windowed signal segment. This process effectively reduces spectral leakage caused by signal truncation and improves the accuracy of spectral analysis.

[0070] Step A42: Perform a short-time Fourier transform on each windowed signal segment to obtain the transformed data; It should be noted that the Short-Time Fourier Transform (STFT) is a time-frequency analysis method that segments a signal and performs Fourier transforms on it to observe how the signal frequency changes over time. The transformed data refers to the complex frequency domain representation obtained after the STFT, which includes amplitude and phase information.

[0071] Understandably, this step performs a short-time Fourier transform on each windowed signal segment to obtain the transformed data, which reflects the frequency component distribution of the signal in each time period, providing a basis for subsequent energy spectrum calculations.

[0072] Step A43: Squaring the real part of the transformed data and calculating the energy spectrum; It should be noted that squaring the real part refers to squaring the real part of a complex number to extract the actual energy component of the signal. The energy spectrum refers to the spectral lines representing the energy distribution of a signal at various frequency points, and is an important feature in time-frequency analysis.

[0073] Understandably, this step involves squaring the real part of the transformed data and calculating the energy spectrum. The squaring operation eliminates the influence of phase information, highlights the energy distribution characteristics of the signal, and provides a data basis for subsequent normalization processing.

[0074] Step A44: Perform nonlinear normalization on the energy spectrum to obtain the intensity spectrum data of the Barker domain; It should be noted that nonlinear normalization refers to using a nonlinear transformation method to map data to a uniform scale, thereby eliminating amplitude differences between different signal segments. The Bark domain is a frequency scale based on the characteristics of human hearing, converting linear frequencies into a nonlinear scale that better matches the subjective perception of sound. Intensity spectrum data refers to the signal intensity values ​​represented in the Bark domain after normalization, used for subsequent curve generation.

[0075] Understandably, this step performs nonlinear normalization on the energy spectrum to obtain the intensity spectrum data in the Barker domain. This process not only eliminates the influence of signal amplitude differences but also transforms the frequency scale to the human ear's perception domain, making subsequent analysis closer to the actual listening effect.

[0076] Step A45: Generate an intensity spectrum curve based on the intensity spectrum data.

[0077] It should be noted that the intensity spectrum curve is a continuous curve with Barker domain or frequency as the horizontal axis and signal intensity as the vertical axis, which intuitively shows the intensity distribution characteristics of the signal in different frequency bands.

[0078] Understandably, this step generates intensity spectrum curves based on intensity spectrum data, connecting discrete data points into a continuous curve for easy visual observation and subsequent comparative analysis.

[0079] Step S202: Perform a Fast Fourier Transform on the digital signal to obtain the spectrum curve; It should be noted that the Fast Fourier Transform (FFT) is a highly efficient discrete Fourier transform algorithm used to quickly convert time-domain signals into frequency-domain signals. A frequency spectrum curve is a continuous curve plotted on the x-axis with signal amplitude on the y-axis, reflecting the energy distribution of the signal across its various frequency components.

[0080] Understandably, this step performs a Fast Fourier Transform on the digital signal to obtain the spectral curve. This transformation yields the global frequency characteristics of the signal, providing frequency domain information for subsequent multi-dimensional feature analysis.

[0081] In practical implementation, FFT spectrum analysis can help identify the source and characteristics of noise. By performing Fourier Transform (FFT) or Short Time Fourier Transform (STFT) on the signal, a spectrum is obtained. The spectrum can show the energy distribution of different frequency components in the noise, which helps to identify specific noise sources or abnormal frequency components.

[0082] It should be understood that FFT is a fast algorithm for Discrete Fourier Transform, which can transform a signal to the frequency domain. Some signals are difficult to identify in the time domain, but their characteristics become readily apparent after transformation to the frequency domain. Furthermore, FFT can extract the spectrum of a signal, which is frequently used in spectral analysis. An analog signal, after being sampled by an ADC, becomes a digital signal. According to the Nayguister's law, the sampling frequency must be greater than twice the signal frequency. The sampled digital signal can then undergo FFT transformation. N sampling points, after FFT, yield N points of FFT result. For ease of FFT calculation, N is usually taken as an integer power of 2. Assuming the sampling frequency is Fs, the signal frequency is F, and the number of sampling points is N, then the result after FFT is an N-point complex number. Each point corresponds to a frequency point. The magnitude of this point is the amplitude characteristic at that frequency value. The relationship with the amplitude of the original signal is as follows: assuming the peak value of the original signal is A, then the magnitude of each point in the FFT result (except for the DC component at the first point) is N / 2 times A. The first point is the DC component, and its magnitude is N times that of the DC component. After the FFT, a point n is represented by the complex number a + bi, then the magnitude of this complex number is An = √(a^2 + b^2), and the phase is Pn = atan2(b, a).

[0083] Step S203: Perform characteristic loudness and sharpness analysis on the digital signal to obtain loudness curve and sharpness curve; It should be noted that characteristic loudness and sharpness analysis refers to the analytical method of calculating the loudness value perceived by the human ear and the sharpness value reflecting the stimulating effect of high-frequency components based on the Zwicker psychoacoustic model. A loudness curve is a continuous curve with the Bark domain as the x-axis and the loudness value as the y-axis, representing the distribution of sound volume perceived by the human ear as a function of frequency. A sharpness curve is a continuous curve with the Bark domain as the x-axis and the sharpness value as the y-axis, reflecting the subjective feeling of high-frequency components in a sound.

[0084] Understandably, this step performs characteristic loudness and sharpness analysis on the digital signal to obtain loudness and sharpness curves. This analysis simulates the subjective perception characteristics of sound by the human ear, making the detection standard more consistent with the actual listening experience.

[0085] In one feasible implementation, step S203 may include steps A51 to A58: Step A51: Calculate the sound pressure level of the digital signal in each critical frequency band in the acoustic model to obtain sound pressure level data; It should be noted that the critical band refers to the smallest bandwidth within the human auditory system that allows sound to be distinguished into different frequency components; it is the basic unit of psychoacoustic analysis. Sound pressure level (SPL) refers to the level of sound pressure, used to quantify the physical intensity of sound. SPL data refers to the set of SPL values ​​calculated across various critical bands.

[0086] Understandably, this step calculates the sound pressure level of the digital signal in each critical frequency band within the acoustic model, obtaining sound pressure level data. By decomposing the signal into different critical frequency bands, it provides a frequency-band data basis for subsequent loudness calculations.

[0087] Step A52: Obtain preset hearing threshold sound pressure level data, and correct the sound pressure level data according to the preset hearing threshold sound pressure level data to obtain the corrected sound pressure level; It should be noted that the hearing threshold sound pressure level (HLP) refers to the minimum sound pressure level that the human ear can perceive in a quiet state, and it is an important reference benchmark in psychoacoustic models. The corrected sound pressure level is the value obtained by subtracting the hearing threshold sound pressure level from the actual sound pressure level, reflecting the actual effective sound intensity exceeding the hearing threshold.

[0088] Understandably, this step obtains preset hearing threshold sound pressure level data, corrects the sound pressure level data based on the preset hearing threshold sound pressure level data, and obtains the corrected sound pressure level. This correction process excludes sound components that cannot be perceived by the human ear, making subsequent calculations more consistent with the actual hearing characteristics of the human ear.

[0089] Step A53: Calculate the characteristic loudness value of each frequency band based on the corrected sound pressure level and the preset characteristic loudness formula to obtain the characteristic loudness data; It should be noted that characteristic loudness values ​​refer to the loudness components perceived by the human ear within each critical frequency band, and are intermediate parameters for calculating total loudness. Characteristic loudness data refers to the set of characteristic loudness values ​​for each frequency band, reflecting the distribution of loudness across different frequency bands.

[0090] Understandably, this step calculates the characteristic loudness value of each frequency band based on the corrected sound pressure level and the preset characteristic loudness formula, and obtains characteristic loudness data. This calculation converts the physical sound pressure level into psychoacoustic loudness perception, providing frequency band data for subsequent integration calculations.

[0091] Step A54: Integrate the characteristic loudness data in the Barker domain to obtain the total loudness value; It should be noted that the total loudness value refers to the total loudness perceived by the human ear across the entire range of auditory frequencies, and is a comprehensive indicator for evaluating the loudness of a sound.

[0092] Understandably, this step involves integrating the characteristic loudness data in the Barker domain to obtain the total loudness value. This integration process sums the characteristic loudness components of each frequency band to obtain the overall loudness perceived by the human ear, providing crucial data for subsequent curve generation.

[0093] In practical implementation, during loudness analysis and noise testing, the difference between normal and abnormal responses from the speaker is noticeable to the human ear. Therefore, the difference can be artificially distinguished based on auditory perception. Based on this, loudness or sharpness can be used for analysis. Loudness is the subjective perception of sound intensity; it is a concept in psychoacoustics, usually measured in phons or sones. Loudness typically increases with sound pressure level (SPL), but the relationship is not linear. The human ear's perception of loudness gradually decreases with increasing SPL. Different frequencies of sound at the same SPL are perceived differently by the human ear. Generally, the human ear is more sensitive to mid-to-high frequency sounds and less sensitive to low-frequency sounds. This system uses the Zwicker characteristic loudness calculation method for analysis, with the following formula:

[0094] Where, N Lp is a special symbol used in Zwicker characteristic loudness calculation to represent characteristic loudness. Lp is the sound pressure level corresponding to the center frequency f under actual conditions; LPTQ is the sound pressure level corresponding to the center frequency f under quiet conditions, that is, the sound pressure level corresponding to each critical center frequency on the equal loudness curve.

[0095] N on 0 to 24 Bark (z) The total loudness N is obtained by integration, as shown in the formula below:

[0096] Step A55: Determine the loudness curve based on the total loudness value; Understandably, this step determines the loudness curve based on the total loudness value, and generates a continuous curve by matching the total loudness value with each frequency band, which intuitively shows the distribution characteristics of loudness in frequency.

[0097] Step A56: Obtain the preset sharpness weighting function, and perform weighted calculation on the characteristic loudness data according to the preset sharpness weighting function to obtain the weighted loudness value of each frequency band; It should be noted that the sharpness weighting function refers to the loudness weighting coefficients corresponding to different critical frequency bands under the Zwicker model, used to highlight the contribution of high-frequency components to sharpness. The weighted loudness value is the loudness value after being weighted by the weighting function, reflecting the degree of influence of different frequency components on sharpness.

[0098] Understandably, this step obtains a preset sharpness weighting function, and then performs weighted calculations on the characteristic loudness data based on the preset sharpness weighting function to obtain the weighted loudness value of each frequency band. This weighting process strengthens the influence of high-frequency components, making the sharpness calculation more consistent with the subjective perception of high-frequency noise by the human ear.

[0099] Step A57: Integrate the weighted loudness value in the Barker domain to obtain the sharpness value; It should be noted that the sharpness value is a comprehensive indicator reflecting the perceived value of high-frequency components in a sound, and its unit is acum. It is directly related to the noise spectrum structure.

[0100] Understandably, this step involves integrating the weighted loudness value in the Barker domain to obtain the sharpness value. This integration process accumulates the weighted loudness components of each frequency band to obtain a comprehensive index for evaluating the sharpness of the sound.

[0101] In practical implementation, sharpness corresponds to the perceived value of high-frequency components in noise, and is directly related to the structure of the noise spectrum. The unit of sharpness is acum. Using the Zwicker model, the calculation formula is as follows:

[0102] In the formula, g(z) represents the loudness weighting function corresponding to different critical frequency bands under the Zwicker model.

[0103] Please refer to the formula below for details:

[0104] Step A58: Determine the sharpness curve based on the sharpness value.

[0105] Understandably, this step determines the sharpness curve based on the sharpness value, and generates a continuous curve by matching the sharpness value with each frequency band, which intuitively displays the sharpness characteristics of the sound.

[0106] Step S204: Use the intensity spectrum curve, the frequency spectrum curve, the loudness curve, and the sharpness curve as characteristic curves.

[0107] Understandably, this step unifies the intensity spectrum curve, frequency spectrum curve, loudness curve, and sharpness curve as characteristic curves. These multi-dimensional curves together constitute the complete acoustic feature representation of the product, providing a comprehensive data foundation for subsequent comparison and judgment.

[0108] In this embodiment, the acoustic model automatically performs time-domain feature transformation and multi-curve comparison analysis, which significantly reduces the false negative and false positive rates of manual listening detection. Standardized testing can be achieved without professional training, greatly reducing equipment and personnel training costs. At the same time, it supports multi-channel parallel testing, which effectively improves the efficiency of production line testing and prevents defective products from entering the market.

[0109] It should be noted that the above examples are only for understanding this application and do not constitute a limitation on the noise detection method for audio products in this application. Any simple modifications based on this technical concept are within the protection scope of this application.

[0110] This application also provides an audio product noise detection device; please refer to... Figure 5 The noise detection device for audio products includes: Acquisition module 10 is used to receive the sound signal of the audio product under test and convert the sound signal into a digital signal; Feature extraction module 20 is used to perform time-domain feature transformation on digital signals through an acoustic model to obtain feature curves; The comparison module 30 is used to compare each feature curve with a preset target curve to obtain the comparison results; The judgment module 40 is used to set a lower limit in the target frequency band based on the comparison results; if all characteristic curves are above the lower limit, the audio product under test is judged to be qualified.

[0111] This embodiment provides an audio product noise detection device. This embodiment receives the sound signal of the audio product under test and converts the sound signal into a digital signal; performs time-domain feature transformation on the digital signal through an acoustic model to obtain a feature curve; compares each feature curve with a preset target curve to determine the target frequency band; sets a lower limit threshold in the target frequency band, and determines that the audio product under test is qualified if all feature curves are above the lower limit threshold.

[0112] In summary, this embodiment automatically performs time-domain feature transformation and multi-curve comparison analysis through acoustic models, which significantly reduces the false negative and false positive rates of manual sound detection. It can achieve standardized testing without professional personnel, and supports multi-channel parallel testing, which improves the efficiency of production line testing, effectively prevents defective products from entering the market, and greatly reduces equipment costs and personnel training costs.

[0113] Optionally, the acquisition module 10 is also used to forward the acquisition signal detected by the soundproof box to the host computer so as to control the microphone sensor to pick up the sound signal generated by the audio product under test during operation. The audio signal is transmitted to the sound card to convert it into a digital audio signal.

[0114] Optionally, the feature extraction module 20 is also used to acquire the original sound signals generated during the operation of a preset number of audio products; The sound card converts the original analog audio signal into a digital signal for training. The training digital signal is bandpass filtered to retain the target training signal within a preset frequency range; Custom feature extraction and calibration are performed on the target training signal to obtain the sound feature parameters of the target training signal; By using autocorrelation comparison algorithm, cross-correlation comparison algorithm and signal similarity analysis, sound feature parameters are screened and compared to separate good product feature signals and defective product feature signals; Good product feature signals and defective product feature signals are stored in the feature sample library to expand the training sample size. The similarity and correlation between each training sample and the good product feature signal are compared. The model parameters are iteratively optimized to determine the target signal feature template and obtain the trained acoustic model.

[0115] Optionally, the feature extraction module 20 is also used to input the digital signal into the acoustic model, perform normalized intensity spectrum analysis on the digital signal, and obtain the intensity spectrum curve; Perform a Fast Fourier Transform on the digital signal to obtain its spectrum curve; Characteristic loudness and sharpness analysis is performed on digital signals to obtain loudness curves and sharpness curves; Intensity spectrum curve, frequency spectrum curve, loudness curve, and sharpness curve are used as characteristic curves.

[0116] Optionally, the feature extraction module 20 is also used to segment the digital signal in the acoustic model, apply a window function to each signal segment, and obtain a windowed signal segment; Perform a short-time Fourier transform on each windowed signal segment to obtain the transformed data; Square the real part of the transformed data and calculate the energy spectrum; The energy spectrum is nonlinearly normalized to obtain the intensity spectrum data of the Barker domain; Intensity spectrum curves are generated based on intensity spectrum data.

[0117] Optionally, the feature extraction module 20 is also used to calculate the sound pressure level of the digital signal in each critical frequency band in the acoustic model to obtain sound pressure level data; Obtain preset hearing threshold sound pressure level data, correct the sound pressure level data based on the preset hearing threshold sound pressure level data, and obtain the corrected sound pressure level; The characteristic loudness values ​​of each frequency band are calculated based on the corrected sound pressure level and the preset characteristic loudness formula to obtain the characteristic loudness data; The characteristic loudness data are integrated in the Barker domain to obtain the total loudness value. Determine the loudness curve based on the total loudness value; Obtain a preset sharpness weighting function, and perform weighted calculation on the characteristic loudness data according to the preset sharpness weighting function to obtain the weighted loudness value of each frequency band; The sharpness value is obtained by integrating the weighted loudness value in the Barker domain. The sharpness curve is determined based on the sharpness value.

[0118] Optionally, the comparison module 30 is also used to obtain the reference characteristic curve of the preset gold sample, wherein the reference characteristic curve includes a reference intensity spectrum curve, a reference spectrum curve, a reference loudness curve, and a reference sharpness curve. Calculate the similarity data between each characteristic curve and the benchmark characteristic curve in each corresponding frequency band; The similarity data is compared with a preset similarity threshold to filter out the target frequency band; The determination module 40 is also used to set a lower limit threshold in the target frequency band to obtain the determination rules; According to the judgment rules, each characteristic curve is verified band by band. When each characteristic curve is above the lower limit threshold, the audio product under test is judged to be qualified. When all characteristic curves are below the lower threshold, the audio product under test is deemed unqualified, and the process returns to the step of performing time-domain feature transformation on the digital signal through an acoustic model to obtain the characteristic curves.

[0119] The audio product noise detection device provided in this application, employing the audio product noise detection method in the above embodiments, can ensure the efficient recording of data. Compared with the prior art, the beneficial effects of the audio product noise detection device provided in this application are the same as those of the audio product noise detection method provided in the above embodiments, and other technical features in the audio product noise detection device are the same as those disclosed in the methods of the above embodiments, and will not be repeated here.

[0120] This application provides an audio product noise detection device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the audio product noise detection method in Embodiment 1 above.

[0121] The following is for reference. Figure 6This document illustrates a structural schematic diagram of an audio product noise detection device suitable for implementing embodiments of this application. The audio product noise detection device in this application may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Description), PMPs (Portable Media Players), and in-vehicle terminals (e.g., in-vehicle navigation terminals), as well as fixed terminals such as digital TVs and desktop computers. Figure 6 The audio product noise detection device shown is merely an example and should not be construed as limiting the functionality and scope of use of the embodiments of this application.

[0122] like Figure 6 As shown, the audio product noise detection device may include a processing unit 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1002 or a program loaded from a storage device 1003 into a random access memory (RAM) 1004. The RAM 1004 also stores various programs and data required for the operation of the audio product noise detection device. The processing unit 1001, ROM 1002, and RAM 1004 are interconnected via a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to the I / O interface 1006: input devices 1007 including, for example, a touchscreen, touchpad, keyboard, mouse, image sensor, microphone, accelerometer, gyroscope, etc.; output devices 1008 including, for example, a liquid crystal display (LCD), speaker, vibrator, etc.; storage devices 1003 including, for example, magnetic tape, hard disk, etc.; and communication devices 1009. Communication device 1009 allows the audio product noise detection device to communicate wirelessly or wiredly with other devices to exchange data. Although the figure shows audio product noise detection devices with various systems, it should be understood that implementation or possession of all the systems shown is not required. More or fewer systems may be implemented alternatively.

[0123] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 1003, or installed from ROM 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this application.

[0124] The audio product noise detection device provided in this application, employing the audio product noise detection method in the above embodiments, can ensure efficient data burning. Compared with the prior art, the beneficial effects of the audio product noise detection device provided in this application are the same as those of the audio product noise detection method provided in the above embodiments, and other technical features of this audio product noise detection device are the same as those disclosed in the previous embodiment method, and will not be repeated here.

[0125] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0126] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0127] This application provides a computer-readable storage medium having computer-readable program instructions (i.e., a computer program) stored thereon, the computer-readable program instructions being used to execute the audio product noise detection method in the above embodiments.

[0128] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, system, or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0129] The aforementioned computer-readable storage medium may be included in the audio product noise detection device; or it may exist independently and not assembled into the audio product noise detection device.

[0130] The aforementioned computer-readable storage medium carries one or more programs. When the aforementioned one or more programs are executed by the audio product noise detection device, the audio product noise detection device causes the following: to receive the sound signal of the audio product under test and convert the sound signal into a digital signal; to perform time-domain feature transformation on the digital signal through an acoustic model to obtain a feature curve; to compare each feature curve with a preset target curve to determine the target frequency band; to set a lower limit threshold in the target frequency band; and to determine that the audio product under test is qualified if all feature curves are above the lower limit threshold.

[0131] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a Local Area Network (LAN) or a Wide Area Network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0132] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0133] The modules described in the embodiments of this application can be implemented in software or hardware. The names of the modules do not necessarily limit the functionality of the unit itself.

[0134] The readable storage medium provided in this application is a computer-readable storage medium that stores computer-readable program instructions (i.e., a computer program) for executing the above-described audio product noise detection method. This solves the technical problems of discomfort and limitations caused by long-term use of traditional corrective products. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as those of the audio product noise detection method provided in the above embodiments, and will not be repeated here.

[0135] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the audio product noise detection method described above.

[0136] The computer program product provided in this application can solve the technical problems of discomfort and limitations caused by long-term use of traditional orthodontic products. Compared with the prior art, the beneficial effects of the computer program product provided in this application are the same as those of the audio product noise detection method provided in the above embodiments, and will not be repeated here.

[0137] The above are only some embodiments of this application and do not limit the patent scope of this application. All equivalent structural transformations made under the technical concept of this application and using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included in the patent protection scope of this application.

Claims

1. A method for detecting noise in audio products, characterized in that, The method includes: Receive the sound signal from the audio product under test and convert the sound signal into a digital signal; The digital signal is subjected to time-domain feature transformation using an acoustic model to obtain the feature curve; Each of the aforementioned characteristic curves is compared with a preset target curve to determine the target frequency band; A lower limit threshold is set for the target frequency band. If all the characteristic curves are above the lower limit threshold, the audio product under test is deemed qualified.

2. The method as described in claim 1, characterized in that, The step of receiving the audio signal from the audio product under test and converting the audio signal into a digital signal includes: The soundproof box detects and forwards the acquisition signal emitted by the audio product under test to the host computer, so as to control the microphone sensor to pick up the sound signal generated by the audio product under test during operation; The sound signal is transmitted to the sound card to convert the sound signal into a digital sound signal.

3. The method as described in claim 1, characterized in that, Before the step of performing time-domain feature transformation on the digital signal using an acoustic model to obtain the feature curve, the method further includes: Acquire the raw sound signals generated during the operation of a preset number of audio products; The original audio signal is converted from an analog signal to a digital signal for training using a sound card. The training digital signal is bandpass filtered to retain the target training signal within a preset frequency range; Custom feature extraction and calibration are performed on the target training signal to obtain the sound feature parameters of the target training signal; The sound feature parameters are screened and compared using autocorrelation comparison algorithm, cross-correlation comparison algorithm, and signal similarity analysis to separate the good product feature signal and the defective product feature signal. The good product feature signals and the defective product feature signals are stored in the feature sample library to expand the training sample size. The similarity and correlation between each training sample and the good product feature signals are compared. The model parameters are iteratively optimized to determine the target signal feature template and obtain the trained acoustic model.

4. The method as described in claim 1, characterized in that, The step of performing time-domain feature transformation on the digital signal using an acoustic model to obtain a feature curve includes: The digital signal is input into the acoustic model, and normalized intensity spectrum analysis is performed on the digital signal to obtain the intensity spectrum curve; Perform a Fast Fourier Transform on the digital signal to obtain the spectrum curve; The digital signal is subjected to characteristic loudness and sharpness analysis to obtain loudness curves and sharpness curves; The intensity spectrum curve, the frequency spectrum curve, the loudness curve, and the sharpness curve are used as the characteristic curves.

5. The method as described in claim 4, characterized in that, The step of inputting the digital signal into the acoustic model and performing normalized intensity spectrum analysis on the digital signal to obtain the intensity spectrum curve includes: In the acoustic model, the digital signal is segmented and processed, and a window function is applied to each signal segment to obtain windowed signal segments; Perform a short-time Fourier transform on each of the windowed signal segments to obtain the transformed data; The energy spectrum is calculated by squaring the real part of the transformed data. The energy spectrum is subjected to nonlinear normalization to obtain the intensity spectrum data of the Barker domain; An intensity spectrum curve is generated based on the intensity spectrum data.

6. The method as described in claim 4, characterized in that, The step of performing characteristic loudness and sharpness analysis on the digital signal to obtain loudness curves and sharpness curves includes: The sound pressure level of the digital signal in each critical frequency band is calculated in the acoustic model to obtain sound pressure level data; Acquire preset hearing threshold sound pressure level data, and correct the sound pressure level data according to the preset hearing threshold sound pressure level data to obtain the corrected sound pressure level; The characteristic loudness values ​​of each frequency band are calculated based on the corrected sound pressure level and the preset characteristic loudness formula to obtain characteristic loudness data. The characteristic loudness data is integrated in the Barker domain to obtain the total loudness value; Determine the loudness curve based on the total loudness value; Obtain a preset sharpness weighting function, and perform weighted calculation on the feature loudness data according to the preset sharpness weighting function to obtain the weighted loudness value of each frequency band; The sharpness value is obtained by integrating the weighted loudness value in the Barker domain; A sharpness curve is determined based on the sharpness value.

7. The method as described in claim 1, characterized in that, The step of comparing each of the characteristic curves with a preset target curve to determine the target frequency band includes: Obtain the reference characteristic curve of a preset gold sample, wherein the reference characteristic curve includes a reference intensity spectrum curve, a reference frequency spectrum curve, a reference loudness curve, and a reference sharpness curve; Calculate the similarity data between each of the aforementioned feature curves and the reference feature curve in each corresponding frequency band; The similarity data is compared with a preset similarity threshold to filter out the target frequency band; The step of setting a lower threshold in the target frequency band and determining that the audio product under test is qualified when all the characteristic curves meet the lower threshold includes: A lower threshold is set for the target frequency band to obtain the judgment rule; According to the judgment rule, each of the characteristic curves is verified band by band. When each of the characteristic curves meets the lower limit threshold, the audio product under test is determined to be qualified. If all the characteristic curves are below the lower threshold, the audio product under test is determined to be unqualified, and the process returns to the step of performing time-domain feature transformation on the digital signal through an acoustic model to obtain the characteristic curves.

8. A noise detection device for audio products, characterized in that, The device includes: The acquisition module is used to receive the sound signal of the audio product under test and convert the sound signal into a digital signal; The feature extraction module is used to perform time-domain feature transformation on the digital signal using an acoustic model to obtain feature curves; The comparison module is used to compare each of the feature curves with a preset target curve to obtain the comparison result; The determination module is used to set a lower limit in the target frequency band based on the comparison results; if all the characteristic curves are above the lower limit, the audio product under test is determined to be qualified.

9. An audio product noise detection device, characterized in that, The device includes: a memory, a processor, and an audio product noise detection program stored in the memory and executable on the processor, the audio product noise detection program being configured to implement the audio product noise detection method as described in any one of claims 1 to 7.

10. A storage medium, characterized in that, The storage medium stores an audio product noise detection program, which, when executed by a processor, implements the audio product noise detection method as described in any one of claims 1 to 7.

Images