A method for reducing noise of intelligent motor
By training an active voice noise reduction calculation model and automatically switching noise reduction mode, the problems of environmental recognition and wind noise leakage in motor active noise reduction technology are solved, and efficient noise reduction and sound quality improvement within the hearing frequency range of the human ear are achieved.
Patent Information
- Application Number
- CN202410227961.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-02-29
- Publication Date
- 2024-12-06
- Estimated Expiration
- 2044-02-29
AI Technical Summary
The existing motor active noise reduction technology cannot automatically identify the environmental state within the hearing frequency range of the human ear and switch the noise reduction mode, and there is a serious problem of wind noise leakage, which limits the noise reduction ability and sound quality.
By training the active voice noise reduction calculation model, the computing power of the noise reduction chip and software algorithm are improved, and noise reduction within the hearing frequency range of the human ear is realized, and the noise reduction mode and wind noise detection control controller parameters are automatically switched through the environment, and the noise reduction mode is automatically switched to the environment to control the controller parameters. Reverse sound waves with the same noise waveform and a phase difference of 180 degrees optimize wind noise leakage.
It realizes efficient noise reduction within the hearing frequency range of the human ear, improves noise reduction capability and sound quality, and optimizes wind noise leakage and enhances the noise reduction performance of the motor.
Smart Images

Figure CN118098189B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of motor noise reduction, and in particular to an intelligent motor noise reduction method. Background Art
[0002] Motors include electric motors and generators. They are devices that realize the mutual conversion between electrical energy and mechanical energy. They have been widely used in various industries. When motors move at high speeds, noise pollution becomes more and more prominent. These noises are not good for workers who have long-term close contact with the equipment. At the same time, strong noises can cause acoustic fatigue of machines, equipment and certain industrial structures. Long-term effects will shorten their service life and even lead to accidents.
[0003] At present, noise reduction methods can be mainly divided into active noise reduction and passive noise reduction. In the field of motors, active noise reduction (ANC) technology is often used. By collecting the noise during motor movement, it regenerates an anti-noise signal with the same amplitude and opposite phase to cancel the noise signal and achieve the purpose of noise reduction. However, the existing motor active noise reduction technology algorithm cannot automatically identify the environmental state within the human hearing frequency range and automatically switch the noise reduction mode according to the environment. In addition, there is a serious problem of wind noise leakage, which limits the noise reduction ability and quality. Summary of the invention
[0004] The purpose of the present invention is to overcome the shortcomings of the prior art and propose an intelligent motor noise reduction method. Through an active voice noise reduction calculation model, the chip computing power and software algorithm are improved to achieve the iteration of noise reduction capability and sound quality. It can also automatically switch between different noise reduction modes according to the environment to optimize wind noise leakage.
[0005] The purpose of the present invention is achieved through the following technical solutions: an intelligent motor noise reduction method, the method is to train an active noise reduction model to obtain an active voice noise reduction calculation model that can reduce noise within the human ear hearing frequency range, and the active voice noise reduction calculation model is set in the noise reduction chip of the motor. The noise reduction chip automatically switches to different noise reduction modes according to the environment, and adjusts its controller parameters according to the wind noise detection result to optimize wind noise leakage, and then generates a reverse sound wave that is the same as the noise waveform and has a phase difference of 180 degrees, so that the collected motor noise sound wave is combined with the reverse sound wave, and finally the motor noise is reduced.
[0006] Furthermore, the active speech noise reduction calculation model includes:
[0007] A secondary channel filtering module is used to filter the initial reference noise signal x(n) and generate a filtered signal x'(n) to the adaptive filtering module;
[0008] The adaptive filtering module calculates the vector factor W according to the filtered signal x'(n) and the error signal e(n) between the noise reduction signal y(n) and the noise signal d(n), and feeds the vector factor W back to the primary channel filtering module;
[0009] The primary channel filtering module filters the initial reference noise signal x(n) in combination with the vector factor W and outputs the noise reduction signal y(n);
[0010] The error sampling value calculation module cancels the noise reduction signal y(n) and the noise signal d(n), calculates the error signal e(n) and outputs it to the adaptive filtering module.
[0011] Further, the adaptive filtering module performs the following operations:
[0012] According to the filtered signal x'(n), iterative operation is performed to calculate the vector factor W, as shown in the following formula:
[0013] W=(1-alpha×Leakage)×W+alpha×e(n)×Xfilterd / energy;
[0014] Among them, Xfilterd is the parameter input vector after filtering, that is, the real-time value of each iteration, which is a constant; alpha is the step factor, Leakage is the leakage factor, e(n) is the error signal, that is, the erroneous sampling frame of each sampling, and energy is the square value of the reference vector after filtering, where the reference vector is the standard value of the preset filtering effect.
[0015] Furthermore, the training of the active noise reduction model to obtain an active speech noise reduction calculation model capable of reducing noise within the human hearing frequency range includes the following steps:
[0016] S1. Determine a reference frequency band, and train an active noise reduction model based on the reference frequency band to obtain a standardized model of the reference frequency band that meets preset noise requirements;
[0017] S2, using a collection device to collect raw data, and preprocessing the raw data to obtain sampled data;
[0018] S3, sending the sampled data to the active noise reduction model, and adjusting the characteristic parameters of the active noise reduction model through feedback output until the evaluation index parameters of the model meet the preset requirements compared with the evaluation index parameters of the standardized model in step S1;
[0019] S4, dividing the human hearing frequency range of 20 Hz-20000 Hz into several frequency bands, performing the operations of steps S1-S3 on the several frequency bands respectively, and completing the model parameter adjustment of each frequency band one by one in the order from low to high frequency bands;
[0020] S5. Splice the several frequency band models obtained in step S4 into an active speech noise reduction calculation model covering 20 Hz-20000 Hz, which is suitable for various environments.
[0021] Further, the step S1 comprises:
[0022] After determining the reference frequency band, the active noise reduction model is trained based on the reference frequency band. The training data is standard data that meets the preset noise reduction requirements, artificial data that meets the standard requirements after digital filtering, or a public standard data set. After training, a standardized model of the reference frequency band that meets the noise requirements is obtained.
[0023] Further, the step S2 comprises:
[0024] The acquisition equipment includes a microphone and a pickup, the preprocessing includes noise elimination and signal enhancement; the feature extraction processing includes normalization, data dimension reduction and data splicing.
[0025] Further, the step S3 comprises:
[0026] The sampling data is sent to the active noise reduction model, and the characteristic parameters of the active noise reduction model are adjusted through feedback output, and the noise reduction distortion and frequency response speed of the model and the standardized model are calculated at this time. The noise reduction distortion and frequency response speed of the model are used as evaluation index parameters, and evaluation weights are set for each evaluation index parameter. The evaluation index parameters of the model are compared with the evaluation index parameters of the standardized model according to the weight from high to low; if the comparison result meets the preset requirements, it is considered that the characteristic parameter adjustment of the model is completed at this time, and step S4 is executed; if the comparison result does not meet the preset requirements, the characteristic parameters of the active noise reduction model are continued to be adjusted through feedback output until they meet the preset requirements.
[0027] A noise reduction chip is used to implement the above-mentioned intelligent motor noise reduction method.
[0028] An intelligent motor comprises the above-mentioned noise reduction chip.
[0029] Compared with the prior art, the present invention has the following advantages and beneficial effects:
[0030] The present invention can improve the computing power of the noise reduction chip and the software algorithm by training the active voice noise reduction calculation model, and can automatically switch different noise reduction modes according to the environment to achieve the iteration of noise reduction capability and sound quality; at the same time, the active voice noise reduction calculation model can adaptively adjust the ANC control parameters according to the wind noise detection results, thereby optimizing the motor wind noise leakage. BRIEF DESCRIPTION OF THE DRAWINGS
[0031] Figure 1 This is an architecture diagram of the active speech noise reduction computing model described in the present invention.
[0032] Figure 2 This is a training flow chart of the active speech noise reduction computing model described in the present invention. DETAILED DESCRIPTION
[0033] The present invention will be further described below in conjunction with specific embodiments.
[0034] Example 1
[0035] The intelligent motor noise reduction method provided in this embodiment trains an active noise reduction model to obtain an active voice noise reduction calculation model that can reduce noise within the human hearing frequency range, and sets the active voice noise reduction calculation model in the noise reduction chip of the motor. The noise reduction chip automatically switches to different noise reduction modes according to the environment, and adjusts its controller parameters according to the wind noise detection result to optimize wind noise leakage, and then generates a reverse sound wave that is the same as the noise waveform and has a phase difference of 180 degrees, so that the collected motor noise sound wave is combined with the reverse sound wave, and finally the motor noise is reduced.
[0036] See also Figure 1 As shown, the active speech noise reduction calculation model includes:
[0037] A secondary channel filtering module is used to filter the initial reference noise signal x(n) and generate a filtered signal x'(n) to the adaptive filtering module;
[0038] The adaptive filtering module calculates the vector factor W based on the filtered signal x'(n) and the error signal e(n) between the noise reduction signal y(n) and the noise signal d(n), feeds the vector factor W back to the primary channel filtering module, and performs the following operations:
[0039] According to the filtered signal x'(n), iterative operation is performed to calculate the vector factor W, as shown in the following formula:
[0040] W=(1-alpha×Leakage)×W+alpha×e(n)×Xfilterd / energy;
[0041] Among them, Xfilterd is the parameter input vector after filtering, that is, the real-time value of each iteration, which is a constant; alpha is the step factor, Leakage is the leakage factor, e(n) is the error signal, that is, the erroneous sampling frame of each sampling, and energy is the square value of the reference vector after filtering, where the reference vector is the standard value of the preset filtering effect.
[0042] The primary channel filtering module filters the initial reference noise signal x(n) in combination with the vector factor W and outputs the noise reduction signal y(n);
[0043] The error sampling value calculation module cancels the noise reduction signal y(n) and the noise signal d(n), calculates the error signal e(n) and outputs it to the adaptive filtering module.
[0044] See also Figure 2 As shown, the training of the active noise reduction model to obtain an active speech noise reduction calculation model capable of reducing noise within the human hearing frequency range includes the following steps:
[0045] S1. Determine a reference frequency band, and train an active noise reduction model based on the reference frequency band to obtain a standardized model of the reference frequency band that meets preset noise requirements, including:
[0046] After determining the reference frequency band, the active noise reduction model is trained based on the reference frequency band. The training data is standard data that meets the preset noise reduction requirements, artificial data that meets the standard requirements after digital filtering, or a public standard data set. After training, a standardized model of the reference frequency band that meets the noise requirements is obtained.
[0047] S2. Use acquisition equipment to collect raw data, and pre-process the raw data to obtain sampled data, including:
[0048] The acquisition device includes a microphone and a pickup, the preprocessing includes noise elimination and signal enhancement; the feature extraction processing includes normalization, data dimension reduction and data splicing;
[0049] Because speech noise has a masking effect on pure tones, the analysis of noise and pure tones requires the use of the critical bandwidth of the speech signal, expressed in Bark. Specifically, a pure tone with a frequency of f can be represented by a band with a center frequency of f and a bandwidth of b. f If the power of the pure tone is equal to the power of the noise band, the pure tone can be heard. The bandwidth b is f0 It is called the critical bandwidth of frequency f, and it continues to shrink b f , the pure sound will become clearer and clearer, based on this, the signal is enhanced to obtain clear sampling data.
[0050] S3, sending the sampled data to the active noise reduction model, and adjusting the characteristic parameters of the active noise reduction model through feedback output until the evaluation index parameters of the model meet the preset requirements compared with the evaluation index parameters of the standardized model in step S1, including:
[0051] The sampling data is sent to the active noise reduction model, and the characteristic parameters of the active noise reduction model are adjusted through feedback output, and the noise reduction distortion and frequency response speed of the model and the standardized model are calculated at this time. The noise reduction distortion and frequency response speed of the model are used as evaluation index parameters, and evaluation weights are set for each evaluation index parameter. The evaluation index parameters of the model are compared with the evaluation index parameters of the standardized model according to the weight from high to low; if the comparison result meets the preset requirements, it is considered that the characteristic parameter adjustment of the model is completed at this time, and step S4 is executed; if the comparison result does not meet the preset requirements, the characteristic parameters of the active noise reduction model are continued to be adjusted through feedback output until they meet the preset requirements; wherein, the preset requirements are artificially set according to the noise reduction requirements required by the actual motor.
[0052] S4, dividing the human hearing frequency range of 20 Hz-20000 Hz into several frequency bands, performing the operations of steps S1-S3 on the several frequency bands respectively, and completing the model parameter adjustment of each frequency band one by one in the order from low to high frequency bands;
[0053] S5. Splice the several frequency band models obtained in step S4 into an active speech noise reduction calculation model covering 20 Hz-20000 Hz, which is suitable for various environments.
[0054] Example 2
[0055] The noise reduction chip provided in this embodiment is used to implement the intelligent motor noise reduction method described in Example 1.
[0056] Example 3
[0057] The intelligent motor provided in this embodiment includes the noise reduction chip described in Embodiment 2.
[0058] The embodiments described above are only preferred embodiments of the present invention and are not intended to limit the scope of implementation of the present invention. Therefore, all changes made according to the shape and principle of the present invention should be included in the protection scope of the present invention.
Claims
1. An intelligent motor noise reduction method, characterized in that: The method is to train an active noise reduction model to obtain an active voice noise reduction calculation model that can reduce noise within the human ear hearing frequency range, and set the active voice noise reduction calculation model in a noise reduction chip of a motor. The noise reduction chip automatically switches different noise reduction modes according to the environment, and adjusts its controller parameters according to the wind noise detection result to optimize wind noise leakage, and then generates a reverse sound wave that is the same as the noise waveform and has a phase difference of 180 degrees, so that the collected motor noise sound wave is combined with the reverse sound wave, and finally the motor noise is reduced. The training of the active noise reduction model to obtain an active voice noise reduction calculation model that can reduce noise within the human ear hearing frequency range includes the following steps: S1. Determine a reference frequency band, and train an active noise reduction model based on the reference frequency band to obtain a standardized model of the reference frequency band that meets preset noise requirements; S2, using a collection device to collect raw data, and performing data preprocessing and feature extraction on the raw data to obtain sampled data; S3, sending the sampled data to the active noise reduction model, and adjusting the characteristic parameters of the active noise reduction model through feedback output until the evaluation index parameters of the model meet the preset requirements compared with the evaluation index parameters of the standardized model in step S1; S4, dividing the human hearing frequency range of 20 Hz-20000 Hz into several frequency bands, performing the operations of steps S1-S3 on the several frequency bands respectively, and completing the model parameter adjustment of each frequency band one by one in the order of the frequency bands from low to high; S5. Splice the several frequency band models obtained in step S4 into an active speech noise reduction calculation model covering 20 Hz-20000 Hz, which is suitable for various environments.
2. The intelligent motor noise reduction method according to claim 1, characterized in that: The active speech noise reduction calculation model includes: A secondary channel filtering module is used to filter the initial reference noise signal x(n) and generate a filtered signal x'(n) to the adaptive filtering module; The adaptive filtering module calculates the vector factor W according to the filtered signal x'(n) and the error signal e(n) between the noise reduction signal y(n) and the noise signal d(n), and feeds the vector factor W back to the primary channel filtering module; The primary channel filtering module filters the initial reference noise signal x(n) in combination with the vector factor W and outputs the noise reduction signal y(n); The error sampling value calculation module cancels the noise reduction signal y(n) and the noise signal d(n), calculates the error signal e(n) and outputs it to the adaptive filtering module.
3. According to the intelligent motor noise reduction method according to claim 2, it is characterized in that: The adaptive filtering module performs the following operations: According to the filtered signal x'(n), iterative operation is performed to calculate the vector factor W, as shown in the following formula: W=(1-alpha×Leakage)×W+alpha×e(n)×Xfilterd / energy; Among them, Xfilterd is the parameter input vector after filtering, that is, the real-time value of each iteration, which is a constant; alpha is the step factor, Leakage is the leakage factor, e(n) is the error signal, that is, the erroneous sampling frame of each sampling, and energy is the square value of the reference vector after filtering, where the reference vector is the standard value of the preset filtering effect.
4. The intelligent motor noise reduction method according to claim 1, characterized in that: The step S1 comprises: After determining the reference frequency band, the active noise reduction model is trained based on the reference frequency band. The training data is standard data that meets the preset noise reduction requirements, artificial data that meets the standard requirements after digital filtering, or a public standard data set. After training, a standardized model of the reference frequency band that meets the noise requirements is obtained.
5. The intelligent motor noise reduction method according to claim 1, characterized in that: The step S2 comprises: The acquisition equipment includes a microphone and a pickup, the preprocessing includes noise elimination and signal enhancement; the feature extraction processing includes normalization, data dimension reduction and data splicing.
6. The intelligent motor noise reduction method according to claim 1, characterized in that: The step S3 comprises: The sampling data is sent to the active noise reduction model, and the characteristic parameters of the active noise reduction model are adjusted through feedback output, and the noise reduction distortion and frequency response speed of the model and the standardized model are calculated at this time. The noise reduction distortion and frequency response speed of the model are used as evaluation index parameters, and evaluation weights are set for each evaluation index parameter. The evaluation index parameters of the model are compared with the evaluation index parameters of the standardized model according to the weight from high to low; if the comparison result meets the preset requirements, it is considered that the characteristic parameter adjustment of the model is completed at this time, and step S4 is executed; if the comparison result does not meet the preset requirements, the characteristic parameters of the active noise reduction model are continued to be adjusted through feedback output until they meet the preset requirements.
7. A noise reduction chip, characterized in that: Used to implement the intelligent motor noise reduction method described in any one of claims 1-6.
8. An intelligent motor, characterized in that: Including the noise reduction chip as claimed in claim 7.
