A weighted fusion wind noise reduction method applied to a helmet

Abstract

This invention proposes a weighted fusion wind noise reduction method for helmets. The method includes: Step 1: Placing reference microphone 1 inside the helmet, slightly to the right of the chin; reference microphone 2 inside the helmet, at the level of the right ear; reference microphone 3 inside the helmet, at the level of the right temple; and an error microphone at the cochlea of ​​the right ear. The signals picked up by multiple reference microphones are weighted and fused into a single reference signal for feedforward active noise control. Step 2: The reference signal, after being trained by a fixed-coefficient feedforward control filter in Step 1, is superimposed with the feedback reference signal, which is iteratively updated in real time, to jointly act on the active noise control. This invention proposes a physical layout of reference microphones for helmet wind noise reduction and a weighted fusion wind noise reduction method for helmets. It achieves good noise reduction performance with low computational load, low hardware resource requirements, and strong practicality.

Description

Technical Field

[0001] This invention relates to a weighted fusion wind noise reduction method for helmets, and to the use of active noise cancellation (ANC) technology in helmets. Background Technology

[0002] With the rapid development of science and technology, while enjoying the convenience brought by technology, people have gradually discovered some drawbacks of development, one of which is noise pollution. As noise pollution has a deeper impact on people's daily lives, the demand for reducing noise interference is also increasing. Traditional noise control belongs to passive noise control (PNC), which has the limitation of being unable to reduce low-frequency noise. The emergence and development of active noise control (ANC) has made up for this deficiency. In recent years, active noise control has gradually developed into one of the main research directions in the field of noise control.

[0003] Wind noise is a major noise source for motorcycles, especially at high speeds. It not only affects rider comfort but can also damage hearing and interfere with communication with other road users. Therefore, reducing wind noise is crucial for improving rider comfort and driving safety. In recent years, many technologies and methods have been proposed to address this problem. Among them, active noise control technology has gained popularity due to its high efficiency and compact size. This technology primarily uses microphones to capture wind noise, analyzes it using active noise control algorithms, and generates a canceling sound wave output to counteract the original wind noise. However, achieving effective active noise control is not simple; it requires high-speed, efficient algorithms and precise sound capture and output equipment. Furthermore, the shape, material, and wind speed of the helmet also affect the noise control effect. In summary, active noise control for motorcycle helmet wind noise reduction is a promising technological direction. Although some technical and practical challenges remain, with further research, it is expected to provide motorcycle riders with a quieter and more comfortable riding experience in the future.

[0004] In existing research on active noise control helmet wind noise reduction, proposed algorithms such as multi-reference noise reduction, coherent weighted fusion noise reduction, and real-time updated gain multi-reference noise reduction have drawbacks such as high computational load and high hardware performance requirements. The present invention proposes a weighted fusion wind noise reduction method for helmets, which aims to address the trade-off between computational load and noise reduction performance. Compared with multi-reference noise reduction, coherent weighted fusion noise reduction, and real-time updated gain multi-reference noise reduction, this method significantly reduces the computational load while maintaining good noise reduction performance, and is highly practical. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of existing algorithms for wind noise reduction in motorcycle helmets, such as high computational load and demanding hardware performance. To balance computational load and noise reduction performance, this invention proposes a weighted fusion wind noise reduction method for helmets, comprising the following steps:

[0006] Step 1: Place reference microphone 1 inside the helmet, slightly to the right of the chin (see attached image). Figure 1 (a) Position 1), with reference microphone 2 placed inside the helmet at the level of the right ear (see attached) Figure 1 (a) Position 2), referencing microphone 3 placed inside the helmet at the level of the right temple (see attached) Figure 1 (a) Position 3), the error microphone is placed in the right cochlea (attached) Figure 1 (a) Position 4), the multiple reference signals x1(n), x2(n), ..., xn picked up by multiple reference microphones are... N (n) are weighted and fused according to power to form a single reference signal x(n), which is then subjected to feedforward active noise control. According to the feedforward FxLMS algorithm, the reference signal x(n) passes through an analog secondary path filter. The output of the filter W and the error signal e(n) received by the error microphone work together to update the feedforward control filter W. f Finally, a stable state is reached, and the feedforward control filter W in the stable state is stored. f The coefficient;

[0007] Step 2: Collect the multiple reference signals x1(n), x2(n), ..., x from the multiple reference microphones. N (n) are weighted and fused according to power to form a single reference signal x(n), which is then used for front-feedback hybrid active noise control, where the feedforward control filter W f The coefficients are the feedforward control filter W saved in step one. f The fixed coefficients, according to the feedback FxLMS algorithm, are used to feed back the reference signal x. b (n) Through simulated secondary path filter The output of the filter W and the error signal e(n) received by the error microphone work together to update the feedback control filter W. b The reference signal x(n) passes through the feedforward control filter W f With feedback reference signal x b (n) Passed through feedback control filter W b The combined effect of these technologies enables front-feedback hybrid active noise control at the target location.

[0008] Furthermore, in step two, the feedback reference signal x b(n) is:

[0009] Through formula The feedback reference signal is obtained, where v(n) is the output of the secondary sound source. Let e(n) represent the analog secondary path filter, and let e(n) represent the error signal.

[0010] Furthermore, in step one, the feedforward control filter W f The update process is as follows:

[0011] Through formula The weight coefficient vector is updated, where W f (n) represents the weight coefficient vector before the feedforward control filter is updated, W f (n+1) represents the updated weight coefficient vector, and x(n) represents the reference signal. Let e(n) represent the analog secondary path filter, and let μ represent the error signal. f This represents the convergence factor for updating the feedforward control filter.

[0012] Furthermore, in step two, the feedback control filter W b The update process is as follows:

[0013] Through formula The weight coefficient vector is updated, where W b (n) represents the weight coefficient vector of the feedback control filter before update, W b (n+1) represents the updated weight coefficient vector, x b (n) represents the feedback reference signal. Let e(n) represent the analog secondary path filter, and let μ represent the error signal. b This represents the convergence factor for updating the feedback control filter.

[0014] In summary, due to the adoption of the above technical solutions, the beneficial effects of this invention are: it greatly reduces the amount of computation while maintaining good noise reduction performance, and it is highly practical. Attached Figure Description

[0015] Figure 1 Here are a system schematic diagram and process flowchart of a weighted fusion wind noise reduction method for helmets according to the present invention: (a) Schematic diagram of helmet wind noise reduction system; (b) Process flowchart of step one; (c) Process flowchart of step two.

[0016] Figure 2 This is a secondary path diagram of the present invention in a specific implementation.

[0017] Figure 3 This is a diagram illustrating the noise reduction performance of the present invention in a specific implementation. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the embodiments and accompanying drawings.

[0019] To address the shortcomings of existing algorithms for wind noise reduction in motorcycle helmets, such as high computational load and demanding hardware performance, this invention proposes a weighted fusion wind noise reduction method for helmets, balancing computational load and noise reduction performance.

[0020] Appendix Figure 1 (a) shows a schematic diagram of the system of the present invention. In this specific embodiment, the reference microphone 1 is placed inside the helmet, slightly to the right of the chin (see attached diagram). Figure 1 (a) Position 1), with reference microphone 2 placed inside the helmet at the level of the right ear (see attached) Figure 1 (a) Position 2), referencing microphone 3 placed inside the helmet at the level of the right temple (see attached) Figure 1 (a) Position 3), the error microphone is placed in the right cochlea (attached) Figure 1 (a) Position 4).

[0021] Appendix Figure 1 (b) and (c) illustrate the process flowchart of the noise reduction algorithm used in this invention. In this specific embodiment, step one involves performing single-channel feedforward active noise control to obtain the feedforward control filter W under stable noise reduction performance. f The coefficients. First, the reference signals x1(n), x2(n), ..., xn picked up by each reference microphone are... N (n) are weighted and fused according to power to form a single reference signal x(n):

[0022] x(n)=[x1(n)G1(n)+x2(n)G2(n)+…+x N (n)G N (n),x(n-1),…,x(n-L+1)] T

[0023] Where G1(n), G2(n), ..., G N (n) represents the gain corresponding to each reference signal:

[0024]

[0025]

[0026] Using x(n) as the reference signal, single-channel feedforward active noise control is performed. According to the feedforward FxLMS algorithm, the reference signal x(n) passes through an analog secondary path filter. The output of the filter W and the error signal e(n) received by the error microphone work together to update the feedforward control filter W. f Finally, a stable state is reached, and the feedforward control filter W in the stable state is stored. f The coefficients. Among them, the feedforward control filter W... f The updated formula is:

[0027]

[0028] Among them, W f (n) represents the weight coefficient vector before the feedforward control filter is updated, W f (n+1) represents the updated weight coefficient vector, and e(n) is the error signal, which is picked up by the error microphone. To simulate the secondary path filter, μ is obtained by training according to the LMS algorithm. f The convergence factor representing the update of the feedforward control filter is a constant in this specific embodiment.

[0029] In this specific embodiment, step two involves forward feedback hybrid active noise control, wherein the feedforward control filter W... f The coefficients are the feedforward control filter W in the stable noise reduction state saved in step one. f The coefficients of this filter remain constant throughout the noise reduction process. The reference signal x(n) is obtained in the same way as in step one. f (n) is the reference signal x(n) passed through the feedforward control filter W. f Subsequent signals:

[0030] v f (n)=x T (n)*W f (n)

[0031] v b (n) is the feedback reference signal x b (n) Passed through feedback control filter W b Subsequent signals:

[0032] v b (n)=x b T (n)*W b (n)

[0033] v(n) is v f (n) and v b Superposition of (n):

[0034] v(n)=v f (n)+v b (n)

[0035] y(n) is the cancellation signal obtained by v(n) through the actual secondary path S(n):

[0036] y(n)=v T (n)*S(n)

[0037] Feedback reference signal x in the feedback control loop b (n) represents the error signal e(n) minus y′(n), where y′(n) is the sum of v(n) and the analog secondary path filter. The calculated result shows that the signal can simulate y(n):

[0038]

[0039] x b (n)=e(n)-y′(n)

[0040] Based on the feedback FxLMS algorithm, the feedback reference signal x b (n) Through simulated secondary path filter The output of the filter W and the error signal e(n) received by the error microphone work together to update the feedback control filter W. b Feedback control filter W b The updated formula is:

[0041]

[0042] Among them, W b (n) represents the weight coefficient vector of the feedback control filter before update, W b (n+1) represents the updated weight coefficient vector, and e(n) is the error signal, which is picked up by the error microphone. To simulate the secondary path filter, μ is obtained by training according to the LMS algorithm. b The convergence factor representing the update of the feedback control filter is a constant in this specific embodiment.

[0043] The reference signal x(n) passes through a feedforward control filter W with fixed coefficients. f With feedback reference signal x b (n) The feedback control filter W that has been updated in real time. b The combined effect of these technologies enables front-feedback hybrid active noise control at the target location.

[0044] In terms of computational complexity (where S represents the secondary path length, L represents the system control filter length, N represents the number of reference microphones, M represents one multiplication operation, A represents one addition operation, and LM represents L multiplication operations; assuming all system control filter lengths are the same):

[0045] The computational cost of performing step one once is: (2L+S+1)M+(2L+S-1)A

[0046] The computational cost of performing step two once is: (3L+2S+1)M+(3L+2S-2)A

[0047] The specific implementation method involves recording audio first, then simulating the acquired recordings using MATLAB R2016b software to predict the noise reduction amount at the target noise reduction location. The experiment is detailed in the attached document. Figure 1 In the system shown in (a), reference microphone 1, reference microphone 2, and reference microphone 3 are placed at positions 1, 2, and 3, respectively. An error microphone is placed at position 4, and a secondary speaker (using the speaker of a headset) is placed at position 5. A speaker is placed 30cm in front of the helmet. A recording of wind noise at a vehicle speed of 50km / h is played as the primary noise source. At the same time, a fan is placed to blow air to simulate actual wind noise. The recordings collected by the three reference microphones and one error microphone in this scenario are obtained.

[0048] All recordings obtained in this experiment were 10 seconds long. During the simulation, the lengths of the system control filters (including feedforward and feedback control filters) and secondary paths were both set to order 384.

[0049] All the following results were obtained through simulation using MATLAB R2016b software:

[0050] Appendix Figure 2 This is the obtained secondary path filter. The system sampling rate is set to 24kHz, and the impulse response length of the secondary path is 16ms (it becomes stationary after 2ms).

[0051] Appendix Figure 3 The noise reduction method of the present invention is demonstrated to achieve noise reduction effect in the case of motorcycle helmets, with an average noise reduction of 19.10dB and a noise reduction depth of 29dB in the frequency band of 63-500Hz.

[0052] In summary, the weighted fusion wind noise reduction method applied to helmets significantly reduces computational load while maintaining good noise reduction performance, demonstrating strong practicality. The specific embodiments described above illustrate the effectiveness of the method proposed in this invention.

[0053] It should be noted that the above embodiments are not intended to limit the scope of protection of the present invention. Equivalent transformations or substitutions made on the basis of the above technical solutions, as well as several improvements such as extensions based on them, all fall within the scope of protection of the claims of the present invention.

Claims

1. A weighted fusion wind noise reduction method applied to helmets, characterized in that, Includes the following steps: Step 1: Place reference microphone 1 inside the helmet, slightly to the right of the chin; reference microphone 2 inside the helmet, at the level of the right ear; reference microphone 3 inside the helmet, at the level of the right temple; and error microphone in the cochlea of ​​the right ear. Collect the multiple reference signals x1(n), x2(n), ..., x from the multiple reference microphones. N (n) are weighted and fused according to power to form a single reference signal x(n), which is then subjected to feedforward active noise control. According to the feedforward FxLMS algorithm, the reference signal x(n) passes through an analog secondary path filter. The output of the filter W and the error signal e(n) received by the error microphone work together to update the feedforward control filter W. f Finally, a stable state is reached, and the feedforward control filter W in the stable state is stored. f The coefficient; Step 2: Collect the multiple reference signals x1(n), x2(n), ..., x from the multiple reference microphones. N (n) are weighted and fused according to power to form a single reference signal x(n), which is then used for front-feedback hybrid active noise control, where the feedforward control filter W f The coefficients are the feedforward control filter W saved in step one. f The fixed coefficients, according to the feedback FxLMS algorithm, are used to feed back the reference signal x. b (n) Through simulated secondary path filter The output of the filter W and the error signal e(n) received by the error microphone work together to update the feedback control filter W. b The reference signal x(n) passes through the feedforward control filter W f With feedback reference signal x b (n) Passed through feedback control filter W b The combined effect of these technologies enables front-feedback hybrid active noise control at the target location.

2. The weighted fusion wind noise reduction method for helmets according to claim 1, characterized in that, In step two, the feedback reference signal x b (n) is: Through formula The feedback reference signal is obtained, where v(n) is the output of the secondary sound source. Let e(n) represent the analog secondary path filter, and let e(n) represent the error signal.

3. The weighted fusion wind noise reduction method for helmets according to claim 1, characterized in that, In step one, the feedforward control filter W f The update process is as follows: Through formula The weight coefficient vector is updated, where W f (n) represents the weight coefficient vector before the feedforward control filter is updated, W f (n+1) represents the updated weight coefficient vector, and x(n) represents the reference signal. Let e(n) represent the analog secondary path filter, and let μ represent the error signal. f This represents the convergence factor for updating the feedforward control filter.

4. The weighted fusion wind noise reduction method for helmets according to claim 1, characterized in that, In step two, the feedback control filter W b The update process is as follows: Through formula The weight coefficient vector is updated, where W b (n) represents the weight coefficient vector of the feedback control filter before update, W b (n+1) represents the updated weight coefficient vector, x b (n) represents the feedback reference signal. Let e(n) represent the analog secondary path filter, and let μ represent the error signal. b This represents the convergence factor for updating the feedback control filter.

Images