AI enhanced full-spectrum sound law mining and analyzing system
By combining traditional and intelligent noise reduction methods in the sound mining and analysis system, and by comparing and identifying various traditional noise reduction techniques, sound signals that better match the target signal are identified. This solves the problem of poor noise reduction performance in complex environments and improves the quality of analysis results.
Patent Information
- Application Number
- CN202511386460.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-09-26
- Publication Date
- 2026-01-02
AI Technical Summary
Existing sound mining and analysis systems have low noise reduction performance in complex sound environments, which affects sound quality and leads to poor analysis results.
The system employs both traditional and intelligent noise reduction methods to process the acquired sound. It also uses a comparison and identification module to identify sound signals that better match the target signal. Furthermore, it combines various traditional noise reduction techniques, such as time-domain filtering, frequency-domain filtering, adaptive filtering, and wavelet transform, to further improve the noise reduction effect.
By combining multiple noise reduction methods, the quality of the sound signal is improved, thereby enhancing the accuracy and reliability of the analysis results.
Smart Images
Figure CN121260179A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of sound signal processing technology, specifically to an AI-enhanced full-spectrum sound pattern mining and analysis system. Background Technology
[0002] The sound mining and analysis system is a tool that combines signal processing, machine learning, and data analysis techniques. It aims to extract regular features from sound data and achieve functions such as pattern recognition, sentiment analysis, or anomaly detection. It has wide applications in public resource transaction monitoring, customer service quality inspection optimization, medical auxiliary diagnosis, and ecological monitoring.
[0003] The core architecture of a sound mining and analysis system typically includes modules such as data acquisition, preprocessing, feature extraction, model building, and result visualization. Among the preprocessing modules, noise reduction is a crucial step. Current technologies generally employ a single noise reduction method, which yields relatively poor results in complex sound environments, leading to lower sound quality and impacting subsequent analysis. Summary of the Invention
[0004] (a) Technical problems to be solved
[0005] To address the shortcomings of existing technologies, this invention provides an AI-enhanced full-spectrum sound pattern mining and analysis system. It employs both traditional and intelligent noise reduction methods to denoise the acquired sound, then compares and contrasts the processing results to identify the sound that best matches the target signal. Compared to traditional single noise reduction methods, the system's superior noise reduction effect, identified through comparison, improves the quality of the analysis results.
[0006] (II) Technical Solution
[0007] To achieve the above objectives, the present invention provides the following technical solution: an AI-enhanced full-spectrum sound pattern mining and analysis system, comprising:
[0008] The data acquisition module is used to collect external sounds;
[0009] The preprocessing module includes noise reduction, comparison and identification, standardization, and frame windowing. The noise reduction includes two methods: traditional noise reduction and intelligent noise reduction. Traditional noise reduction and intelligent noise reduction are used to reduce the noise of the acquired external sound. Then, the comparison and identification module compares and identifies the sound processed by the two different noise reduction methods, and selects the sound signal that matches the target signal better. The standardization and frame windowing modules then process the sound signal.
[0010] A sound library is used to provide various sound data to provide target signals for the comparison and identification module.
[0011] The feature extraction module is used to extract the time-domain and frequency-domain features of sound;
[0012] The model building module is used to discover the rhythm, pitch, and emotional patterns of sound;
[0013] The results visualization module is used to generate spectrograms, spectral graphs, or three-dimensional time-frequency graphs to help users understand sound patterns.
[0014] The above technical solution employs the following approach: The data acquisition module collects external sounds, supporting multiple sound input types, including speech, ambient noise, and biological sounds. The preprocessing module performs noise reduction, standardization, and frame-by-frame windowing on the acquired sounds. Noise reduction includes both traditional and intelligent methods. Both methods reduce noise in the acquired sounds, and the processing results are compared and evaluated to identify the sounds that best match the target signal. Compared to traditional single noise reduction methods, the improved noise reduction effect, identified through comparison, helps improve the quality of the analysis results.
[0015] Preferably, the data acquisition module acquires audio in the range of 20Hz to 20kHz.
[0016] The above technical solution allows the data acquisition module to collect sound across the entire spectrum, resulting in a wider acquisition range and broader applicability. The hardware of the data acquisition module must possess a high signal-to-noise ratio (SNR>60dB) and low distortion (THD<1%), such as a professional microphone matrix or hearing aid-grade sensors.
[0017] Preferably, the conventional noise reduction includes time-domain filtering and frequency-domain filtering.
[0018] Preferably, the collected sound is processed by time-domain filtering and frequency-domain filtering respectively, and then compared and identified to select the sound signal that matches the target signal better. Then, it is compared and identified with the sound processed by intelligent noise reduction.
[0019] The above technical solution employs two methods: frequency domain filtering converts the time-domain signal into the frequency domain (through Fourier transform), utilizing the difference in frequency spectrum distribution between noise and the target signal for filtering. Time-domain filtering directly smooths or predicts the signal in the time domain, suppressing sudden noise or high-frequency interference. Two traditional noise reduction methods process the acquired sound separately, and the processed results are compared, then compared with the sound processed by intelligent noise reduction. Multiple processing methods result in superior processing quality.
[0020] Preferably, the traditional noise reduction also includes adaptive filtering and wavelet transform noise reduction.
[0021] Preferably, the collected sound is processed by time-domain filtering, frequency-domain filtering, adaptive filtering and wavelet transform noise reduction respectively, and then compared and identified to select the sound signal that matches the target signal better. Then, it is compared and identified with the sound processed by intelligent noise reduction.
[0022] The above technical solutions employ: Adaptive filtering dynamically adjusts filter parameters based on the characteristics of the input signal to achieve noise tracking and suppression. Wavelet transform denoising utilizes the multi-scale decomposition characteristics of wavelet basis functions to decompose the signal into different frequency sub-bands, and suppresses noise sub-bands through thresholding. Multiple traditional denoising methods are used to process the acquired sound separately, and the processed results are compared before being compared with the sound processed by intelligent denoising. These multiple processing methods result in superior processing quality.
[0023] (III) Beneficial Effects
[0024] Compared with existing technologies, this invention provides an AI-enhanced full-spectrum sound pattern mining and analysis system, which has the following beneficial effects:
[0025] 1. This AI-enhanced full-spectrum sound pattern mining and analysis system comprises a data acquisition module, a preprocessing module, and a sound library. The preprocessing module includes both traditional and intelligent noise reduction methods. Both methods denoise the acquired sound, and the processing results are compared and evaluated to identify the sound that best matches the target signal. Compared to traditional single noise reduction methods, the superior noise reduction effect identified through comparison improves the quality of the analysis results.
[0026] 2. This AI-enhanced full-spectrum sound pattern mining and analysis system employs multiple traditional noise reduction methods, including time-domain filtering, frequency-domain filtering, adaptive filtering, and wavelet transform noise reduction. These methods process the acquired sound separately, compare the processed sound, and then compare it with the sound processed by the intelligent noise reduction system. This multi-method approach results in superior processing quality. Attached Figure Description
[0027] Figure 1 This is a flowchart illustrating the overall process of an AI-enhanced full-spectrum sound pattern mining and analysis system proposed in this invention.
[0028] Figure 2 This is a flowchart of the preprocessing module of an embodiment of an AI-enhanced full-spectrum sound pattern mining and analysis system proposed in this invention.
[0029] Figure 3 This is a schematic diagram of the preprocessing module structure of another embodiment of the AI-enhanced full-spectrum sound pattern mining and analysis system proposed in this invention. Detailed Implementation
[0030] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0031] Please see Figure 1-3 An AI-enhanced full-spectrum sound pattern mining and analysis system includes:
[0032] The data acquisition module is used to collect external sounds;
[0033] The preprocessing module includes noise reduction, comparison and discrimination, standardization, and frame windowing. Noise reduction includes two methods: traditional noise reduction and intelligent noise reduction. Traditional noise reduction and intelligent noise reduction are used to reduce the noise of the acquired external sound. Then, the comparison and discrimination module compares and distinguishes the sound processed by the two different noise reduction methods, and selects the sound signal that matches the target signal better. The standardization and frame windowing modules then process the sound signal.
[0034] The sound library provides various sound data to serve as the target signal for the comparison and identification module.
[0035] The feature extraction module is used to extract the time-domain and frequency-domain features of sound;
[0036] The model building module is used to discover the rhythm, pitch, and emotional patterns of sound;
[0037] The results visualization module is used to generate spectrograms, spectral graphs, or three-dimensional time-frequency graphs to help users understand sound patterns.
[0038] The data acquisition module collects external sounds, supporting various sound input types, including speech, ambient noise, and biological sounds. The preprocessing module performs noise reduction, standardization, and frame-by-frame windowing on the acquired sounds. Noise reduction includes both traditional and intelligent methods. Both methods reduce noise in the acquired sounds, and the results are compared and evaluated to identify sounds that better match the target signal. Compared to traditional single-method noise reduction, the comparison identifies superior sound signals, thus improving the quality of the analysis results.
[0039] Example 1: The data acquisition module acquires audio in the range of 20Hz to 20kHz.
[0040] The data acquisition module performs full-spectrum sound acquisition, resulting in a wider acquisition range and broader applicability. The hardware of the data acquisition module needs to have a high signal-to-noise ratio (SNR>60dB) and low distortion (THD<1%), such as a professional microphone matrix or hearing aid-grade sensors.
[0041] Example 2: Traditional noise reduction includes time-domain filtering and frequency-domain filtering. The acquired sound is processed by time-domain filtering and frequency-domain filtering respectively, and then compared and identified to select the sound signal that matches the target signal better. This sound signal is then compared and identified with the sound processed by intelligent noise reduction.
[0042] Frequency domain filtering converts the time-domain signal to the frequency domain (through Fourier transform), utilizing the difference in the frequency spectrum distribution between noise and the target signal for filtering. Time-domain filtering directly smooths or predicts the signal in the time domain, suppressing sudden noise or high-frequency interference. Two traditional noise reduction methods process the acquired sound separately, then compare the processed results with the sound processed by intelligent noise reduction. Multiple processing methods result in superior processing quality.
[0043] Example 3: Traditional noise reduction also includes adaptive filtering and wavelet transform noise reduction. The acquired sound is processed by time-domain filtering, frequency-domain filtering, adaptive filtering, and wavelet transform noise reduction respectively. The sound signal that best matches the target signal is then compared and selected, and then compared and selected with the sound processed by intelligent noise reduction.
[0044] Adaptive filtering dynamically adjusts filter parameters based on the characteristics of the input signal to achieve noise tracking and suppression. Wavelet transform denoising utilizes the multi-scale decomposition characteristics of wavelet basis functions to decompose the signal into different frequency sub-bands, and suppresses noise sub-bands through thresholding. Multiple traditional denoising methods are used to process the acquired sound separately, and the processed results are compared before being compared with the sound processed by intelligent denoising. These multiple processing methods result in superior processing quality.
[0045] In summary, this AI-enhanced full-spectrum sound pattern mining and analysis system, through the establishment of a data acquisition module, a preprocessing module, and a sound library, incorporates both traditional and intelligent noise reduction methods within the preprocessing module. Both methods denoise the acquired sound, and the processing results are then compared and evaluated to identify the sound that best matches the target signal. Compared to traditional single noise reduction methods, the superior noise reduction effect identified through comparison contributes to improving the quality of the analysis results.
[0046] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one" does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.
[0047] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.
Claims
1. An AI-augmented full-spectrum sound regularity mining and analysis system, characterized in that, The application relates to a sound feature extraction method and device. The data acquisition module is used for acquiring external sound; The preprocessing module comprises noise reduction processing, a contrast discrimination module, standardization processing and frame windowing processing, the noise reduction processing comprises traditional noise reduction and intelligent noise reduction, the traditional noise reduction and the intelligent noise reduction are respectively used for performing noise reduction processing on the acquired external sound, then the contrast discrimination module is used for performing contrast discrimination on the sound processed by the two different noise reduction modes, a sound signal more matched with a target signal is selected, and the selected sound signal is processed by the standardization processing module and the frame windowing processing module; The sound library is used for providing various sound data and providing a target signal for the contrast discrimination module; The feature extraction module is used for extracting time domain and frequency domain features of sound; The model construction module is used for mining rhythm, pitch and emotion rules of sound; The result visualization module is used for generating a frequency spectrum, a spectrogram or a three-dimensional time-frequency graph to assist a user in understanding sound rules.
2. The AI-augmented all-spectrum sound regularity mining and analysis system of claim 1, wherein: The data acquisition module acquires audio in a range of 20Hz to 20kHz.
3. The AI-augmented all-spectrum sound regularity mining and analysis system of claim 2, wherein: The traditional noise reduction comprises a time domain filtering method and a frequency domain filtering method.
4. The AI-augmented all-spectrum sound regularity mining and analysis system of claim 3, wherein: The acquired sound is processed by the time domain filtering method and the frequency domain filtering method respectively, contrast discrimination is performed, a sound signal more matched with a target signal is selected, and then contrast discrimination is performed on the sound processed by the intelligent noise reduction.
5. The AI-augmented all-spectrum sound regularity mining and analysis system of claim 3, wherein: The traditional noise reduction also comprises an adaptive filtering method and a wavelet transform noise reduction.
6. The AI-augmented all-frequency sound regularity mining and analysis system of claim 5, wherein: The acquired sound is processed by the time domain filtering method, the frequency domain filtering method, the adaptive filtering method and the wavelet transform noise reduction respectively, contrast discrimination is performed, a sound signal more matched with a target signal is selected, and then contrast discrimination is performed on the sound processed by the intelligent noise reduction.