WO2013067714A1 - 一种降低突发噪音的方法 - Google Patents

一种降低突发噪音的方法 Download PDF

Info

Publication number
WO2013067714A1
WO2013067714A1 PCT/CN2011/082124 CN2011082124W WO2013067714A1 WO 2013067714 A1 WO2013067714 A1 WO 2013067714A1 CN 2011082124 W CN2011082124 W CN 2011082124W WO 2013067714 A1 WO2013067714 A1 WO 2013067714A1
Authority
WO
WIPO (PCT)
Prior art keywords
audio signal
wavelet
audio
burst noise
notch
Prior art date
Application number
PCT/CN2011/082124
Other languages
English (en)
French (fr)
Inventor
吕润春
Original Assignee
Liv Runchun
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Liv Runchun filed Critical Liv Runchun
Priority to PCT/CN2011/082124 priority Critical patent/WO2013067714A1/zh
Publication of WO2013067714A1 publication Critical patent/WO2013067714A1/zh

Links

Classifications

    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L21/00—Speech or voice signal processing techniques to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility
    • G10L21/02—Speech enhancement, e.g. noise reduction or echo cancellation
    • G10L21/0208—Noise filtering
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
    • G10L25/27—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the analysis technique
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L19/00—Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis
    • G10L19/02—Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis using spectral analysis, e.g. transform vocoders or subband vocoders
    • G10L19/0212—Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis using spectral analysis, e.g. transform vocoders or subband vocoders using orthogonal transformation
    • G10L19/0216—Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis using spectral analysis, e.g. transform vocoders or subband vocoders using orthogonal transformation using wavelet decomposition

Definitions

  • the present invention relates to a method of reducing noise, and more particularly to a system and method for reducing noise by a notch filter and wavelet transform analysis. Background technique
  • the collected audio signals include not only the voice of the call but also the ambient noise.
  • the ambient noise and the voice of the person will overlap each other, which seriously affects the quality of the call.
  • only a part of the noise can be removed, and most of the noise is difficult to remove. This is because the voices emitted by people, the frequency bands of certain frequency bands and the environment overlap, and the noise is completely eliminated.
  • the call sounds are eliminated together.
  • a common noise reduction method is to perform noise reduction processing at a specific frequency through a notch filter, but since the audio in communication contains a large amount of audio signals and noise signals, it is not a simple doping with a specific frequency. The audio signal, therefore, the notch filter can only remove parts, especially for high frequency noise signals, and can not achieve satisfactory processing results.
  • Wavelet transform is a local transform of space (time) and frequency, so it can effectively extract information from signals. Multi-scale refinement analysis of functions or signals through computational functions such as scaling and translation. But Wavelet transform analysis is mainly used in image processing, and the wavelet algorithm has high complexity and is not applied to noise processing. Summary of the invention
  • the main object of the present invention is to provide a method for reducing burst noise, which combined with a notch filter can effectively suppress noise and recover a real audio signal, and the method is simple in implementation, high in processing efficiency, and low in cost. Low cost, easy to apply to audio processing in a variety of places.
  • a method of reducing burst noise characterized in that the method comprises the following steps:
  • Notch processing notching the audio signal through the notch filter, filtering out the power frequency interference
  • Wavelet analysis processing wavelet decomposition of the notched audio signal, selecting wavelet and determining the decomposition level as N, according to the Nth layer low frequency coefficient of the wavelet decomposition and the quantized 1 ⁇ N layer high frequency coefficient Wavelet reconstruction
  • the notch processing is to filter out the 50 Hz interference frequency by 50 Hz point resistance filtering.
  • the decomposition level N of the wavelet is up to 6, usually 3-4 layers, which facilitates accurate analysis and reconstruction of the curve of the audio signal.
  • performing wavelet reconstruction refers to performing wavelet calculation in reverse to obtain a new audio signal. Furthermore, the above reconstruction is to first segment the audio signal, the length of each segment is 1024, the last segment is not enough to fill zero, and then the main peak of each segment is used to reconstruct the audio curve, and finally the linear interpolation algorithm is used. To smooth the waveform.
  • the above-mentioned peak reconstructed audio curve first acquires all the peaks of the signal of the segment, then performs sliding smoothing on the non-peak portion, and then reconstructs a trough between the two peaks.
  • the invention adopts the combination of the trap and the wavelet analysis processing to eliminate the noise, and performs wavelet processing on the audio signal, which can effectively eliminate the high frequency noise in the audio signal, and achieve the effect of suppressing noise and recovering the real voice signal.
  • the method is simple to implement noise elimination, high in processing efficiency, and low in cost.
  • FIG. 1 is a schematic diagram of a hardware structure implemented by the present invention.
  • the hardware for implementing the present invention includes an audio acquisition module, a notch filter, a wavelet analysis processing module, a band pass filter, and a gain module.
  • the audio acquisition module collects the audio signal and transmits the audio signal to the trap.
  • the audio acquisition module can be a microphone, a speaker, etc., which only collects data of the audio signal; the trap, usually a digital trap, adaptive trap A type of wave device that notches the audio signal to filter out noise signals of a specific frequency, especially filtering out power frequency interference, and then audio The signal is transmitted to the wavelet analysis processing module.
  • the wavelet analysis processing module is a core module.
  • the wavelet analysis processing module reconstructs the waveform of the audio signal, eliminates high frequency noise in the audio signal, reconstructs the audio signal, and then performs audio output.
  • the implementation of the present invention includes the following steps:
  • the trapper completes the function of blocking, and the audio signal is notch-processed to filter out the power frequency interference.
  • Reconstruction involves the creation and compensation of the audio signal, because the audio signal loses a part of the noise cancellation process, so the audio signal is processed after the noise reduction to compensate for the lost audio signal.
  • the length of each audio signal is taken as 1024, the last segment is not enough to zero, and then the main peak of each segment is used to reconstruct the audio curve. Finally, the linear interpolation algorithm is used to smooth the waveform.
  • the peak reconstructed audio curve is obtained by first acquiring all the peaks of the signal of this segment, then smoothing the non-peak portion, and then reconstructing a trough between the two peaks to form a complete audio curve.
  • Gain processing in order to prevent the audio processing from exceeding the limit, the processed audio signal needs to be processed by the gain module, and the gain module appropriately adjusts the audio of the over-limit portion to meet the output requirement. 5. Output, after gain processing, output the audio signal.
  • the trap and the wavelet analysis processing are organically combined, and the noise in the audio signal, especially the high frequency noise, can be effectively eliminated, and the noise can be effectively suppressed and the real voice signal can be recovered.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computational Linguistics (AREA)
  • Signal Processing (AREA)
  • Health & Medical Sciences (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Human Computer Interaction (AREA)
  • Physics & Mathematics (AREA)
  • Acoustics & Sound (AREA)
  • Multimedia (AREA)
  • Quality & Reliability (AREA)
  • Noise Elimination (AREA)

Abstract

一种降低突发噪音的方法，包括：输入音频信号，采集音频信号输入到陷波器；陷波处理，通过陷波器将音频信号陷波处理，滤掉工频干扰；进行小波分析处理，对陷波处理后的音频信号再进行小波分解，然后再进行小波重构；及音频输出步骤。该方法采用陷波器与小波分析处理结合的方式进行噪音的消除，达到抑制噪声、恢复真实的语音信号的效果。

Description

一种降低突发噪音的方法 技术领域
本发明涉及噪音的降低方法，准确地说是一种陷波滤波器和小波 变换分析来降低噪音的系统及方法。 背景技术
通信设备对音频信号时， 采集的音频信号不仅包括通话声音， 还 包含有环境噪音， 环境噪音和人的通话声音会互相重叠， 严重影响到 通话的质量。而在目前的音频信号处理过程中， 仅仅能够去除一部分 噪音， 大部分噪音难以去掉， 这是因为人所发出的语音， 某些频段和 环境的噪音频段重叠， 要完全消除噪音， 就会把人们的通话声音一起 消除掉。
目前，一个常用的降低噪音的方法是通过陷波滤波器进行特定频 率的降噪处理，但是由于在通信中的音频中包含大量的音频信号和噪 音信号， 并不是一个单纯的掺杂有特定频率的音频信号， 因此， 陷波 滤波器只能去除部分， 特别是对于高频的噪音信号， 无法达到满意的 处理结果。
近年来， 兴起了小波变换分析方法， 小波变换是空间（时间) 和频率的局部变换， 因而能有效地从信号中提取信息。 通过伸缩 和平移等运算功能可对函数或信号进行多尺度的细化分析。 但是 小波变换分析目前主要应用于图像处理中， 且小波算法的复杂成 度高， 并未应用于对噪音的处理。 发明内容
基于此， 本发明的主要目的是提供一种降低突发噪音的方法， 该 方法结合陷波滤波器， 能够有效地抑制噪声、 恢复真实的音频信号， 且该方法实现简单、 处理效率高、 成本低廉， 便于应用到各种场所的 音频处理中。
因此， 本发明是这样实现的：
一种降低突发噪音的方法， 其特征在于该方法包括如下步骤：
1、 输入音频信号， 采集音频信号输入到陷波器；
2、 陷波处理， 通过陷波器将音频信号陷波处理， 滤掉工频干扰;
3、 小波分析处理， 对陷波处理后的音频信号再进行小波分解， 选择小波并确定分解层次为 N， 根据小波分解的第 N层低频系数和 经过量化后的 1〜N层高频系数进行小波重构；
4、 音频输出。
更进一步， 步骤 2中， 陷波处理是通过 50Hz点阻滤波， 过滤掉 50Hz的干扰频率。
更进一步， 步骤 3中， 小波的分解层次 N最高为 6， 通常为 3-4 层， 便于对音频信号的曲线进行准确分析和重构。
更进一步， 进行小波重构是指反向进行小波计算， 得出新的音频 信号。 更进一步， 上述的重构， 是先对音频信号进行分段处理， 每段长 度取为 1024， 最后一段不够补零， 再利用每段分析的主要峰值来重 构音频曲线， 最后使用线性插值算法来平滑波形。
上述的峰值重构音频曲线是先获取本段信号的所有峰值，然后对 非峰值部分进行滑动平滑， 然后在 2个峰值之间重建一个波谷。
本发明采用陷波器与小波分析处理结合的方式进行噪音的消除， 对音频信号进行小波处理，能够有效地对音频信号中的高频噪音进行 消除， 达到抑制噪声、 恢复真实的语音信号的效果， 且该方法对噪声 的消除实现简单、 处理效率高、 成本低廉。 附图说明
图 1为本发明所实施的硬件结构示意图。
图 2为本发明所实施的原理示意图。 具体实施方式
下面， 结合附图所示， 对本发明的具体实施做详细说明。
图 1所示， 在一种具体是实现方式中， 实现本发明的硬件包括音 频采集模块、 陷波器、 小波分析处理模块、 带通滤波器及增益模块。
音频采集模块采集音频信号， 并将音频信号传输给陷波器， 音频 采集模块， 可以是麦克风、 扬声器等， 其只采集音频信号的数据； 陷 波器， 通常为数字陷波器、 自适应陷波器的一种， 对音频信号进行陷 波处理， 滤掉特定频率的噪音信号， 特别是滤除工频干扰， 再将音频 信号传输给小波分析处理模块。
小波分析处理模块是核心模块， 通过小波分析处理模块， 将音频 信号的波形进行重构， 消除音频信号中的高频噪音， 重构音频信号， 再进行音频输出。
结合图 2所示， 本发明的实现包括如下步骤：
1、 输入音频信号， 音频采集模块， 采集音频信号输入到陷波器。
2、 陷波处理， 陷波器完成带阻的功能， 将音频信号陷波处理， 滤掉工频干扰。
3、 小波分析处理， 对音频信号进行小波分解， 选择小波并确定 分解层次为 N， N=4， 然后根据小波分解的第 N层低频系数和经过 量化后的 1〜N层高频系数进行小波重构。
重构包括对音频信号的建立和弥补，因为音频信号在消噪的过程 当中会损失一部分， 因此在降噪后要对音频信号进行处理， 以弥补损 失的音频信号。 取音频信号每段长度取为 1024， 最后一段不够补零， 再利用每段分析的主要峰值来重构音频曲线，最后使用线性插值算法 来平滑波形。
峰值重构音频曲线是先获取本段信号的所有峰值，然后对非峰值 部分进行滑动平滑， 然后在 2个峰值之间重建一个波谷， 形成完整的 音频曲线。
4、 增益处理， 为了防止音频处理后越限， 还需要将处理后的 音频信号经过增益模块进行处理，增益模块把越限部分的音频进 行适当的调整， 达到输出要求。 5、 输出， 增益处理后， 将音频信号进行输出。
采用上述的方式， 将陷波器与小波分析处理有机地结合在一起， 可以有效地对音频信号中的噪音， 特别是高频噪音进行消除， 能够有 效地抑制噪声、 恢复真实的语音信号。
以上所述仅为本发明的优选实施例而已， 并不用以限制本发明， 凡在本发明的精神和原则之内所作的任何修改、 等同替换和改进等， 均应包含在本发明的保护范围之内。

Claims

1、一种降低突发噪音的方法， 其特征在于该方法包括如下步骤：
1 )、 输入音频信号， 采集音频信号输入到陷波器；
2 )、陷波处理，通过陷波器将音频信号陷波处理，滤掉工频干扰；
3 )、 小波分析处理， 对陷波处理后的音频信号再进行小波分解， 选择小波并确定分解层次为 N， 根据小波分解的第 N层低频系数和 经过量化后的 1〜N层高频系数进行小波重构；
4 )、 音频输出。
2、 如权利要求 1所述的降低突发噪音的方法， 其特征在于步骤 2中， 陷波处理是通过 50Hz点阻滤波， 过滤掉 50Hz的干扰频率。
3、 如权利要求 1所述的降低突发噪音的方法， 其特征在于步骤 3中， 小波的分解层次 N最高为 6， 通常为 3-4层， 便于对音频信号 的曲线进行准确分析和重构。
4、 如权利要求 3所述的降低突发噪音的方法， 其特征在于进行 小波重构是指反向进行小波计算， 得出新的音频信号。
5、 如权利要求 4所述的降低突发噪音的方法， 其特征在于上述 的重构, 是先对音频信号进行分段处理， 每段长度取为 1024， 最后 一段不够补零， 再利用每段分析的主要峰值来重构音频曲线， 最后使 用线性插值算法来平滑波形。
6、 如权利要求 5所述的降低突发噪音的方法， 其特征在于上述 的峰值重构音频曲线是先获取本段信号的所有峰值，然后对非峰值部 分进行滑动平滑， 然后在 2个峰值之间重建一个波谷。
PCT/CN2011/082124 2011-11-12 2011-11-12 一种降低突发噪音的方法 WO2013067714A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
PCT/CN2011/082124 WO2013067714A1 (zh) 2011-11-12 2011-11-12 一种降低突发噪音的方法

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2011/082124 WO2013067714A1 (zh) 2011-11-12 2011-11-12 一种降低突发噪音的方法

Publications (1)

Publication Number Publication Date
WO2013067714A1 true WO2013067714A1 (zh) 2013-05-16

Family

ID=48288474

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2011/082124 WO2013067714A1 (zh) 2011-11-12 2011-11-12 一种降低突发噪音的方法

Country Status (1)

Country Link
WO (1) WO2013067714A1 (zh)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104811258A (zh) * 2015-04-03 2015-07-29 深圳邦健生物医疗设备股份有限公司 干扰信号消除方法、装置以及医疗设备
CN107274907A (zh) * 2017-07-03 2017-10-20 北京小鱼在家科技有限公司 双麦克风设备上实现指向性拾音的方法和装置
CN107426197A (zh) * 2017-07-05 2017-12-01 厦门声戎科技有限公司 一种实现隐蔽语音通话的保密通信方法
CN111585663A (zh) * 2020-04-20 2020-08-25 杭州华立电力系统工程有限公司 低压电力线载波通信特征干扰噪声的复现方法
US20220284909A1 (en) * 2019-11-13 2022-09-08 Tencent Music Entertainment Technology (Shenzhen) Co., Ltd. Method, apparatus, and device for transient noise detection

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102176312A (zh) * 2011-01-07 2011-09-07 蔡镇滨 一种通过小波陷波来降低突发噪音的系统及方法

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102176312A (zh) * 2011-01-07 2011-09-07 蔡镇滨 一种通过小波陷波来降低突发噪音的系统及方法

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104811258A (zh) * 2015-04-03 2015-07-29 深圳邦健生物医疗设备股份有限公司 干扰信号消除方法、装置以及医疗设备
CN107274907A (zh) * 2017-07-03 2017-10-20 北京小鱼在家科技有限公司 双麦克风设备上实现指向性拾音的方法和装置
CN107426197A (zh) * 2017-07-05 2017-12-01 厦门声戎科技有限公司 一种实现隐蔽语音通话的保密通信方法
US20220284909A1 (en) * 2019-11-13 2022-09-08 Tencent Music Entertainment Technology (Shenzhen) Co., Ltd. Method, apparatus, and device for transient noise detection
CN111585663A (zh) * 2020-04-20 2020-08-25 杭州华立电力系统工程有限公司 低压电力线载波通信特征干扰噪声的复现方法

Similar Documents

Publication Publication Date Title
US10580430B2 (en) Noise reduction using machine learning
CN105788607B (zh) 应用于双麦克风阵列的语音增强方法
CN102176312B (zh) 一种通过小波陷波来降低突发噪音的系统及方法
CN105741849A (zh) 数字助听器中融合相位估计与人耳听觉特性的语音增强方法
KR20130108063A (ko) 다중 마이크로폰의 견고한 잡음 억제
CN101625869B (zh) 一种基于小波包能量的非空气传导语音增强方法
RU2016101521A (ru) Устройство и способ для генерации адаптивной формы спектра комфотного шума
WO2013067714A1 (zh) 一种降低突发噪音的方法
CN106463106A (zh) 用于音频接收的风噪声降低
JP5595605B2 (ja) 音声信号復元装置および音声信号復元方法
CN102142255B (zh) 一种在音频信号中嵌入及提取数字水印的方法
CN114566176A (zh) 基于深度神经网络的残余回声消除方法及系统
Sultana et al. Performance analysis of adaptive filtering algorithms for denoising of ECG signals
CN103578466A (zh) 基于分数阶傅里叶变换的语音非语音检测方法
JP6316288B2 (ja) 電子透かし埋め込み装置、電子透かし検出装置、電子透かし埋め込み方法、電子透かし検出方法、電子透かし埋め込みプログラム、及び電子透かし検出プログラム
CN103475986A (zh) 基于多分辨率小波的数字助听器语音增强方法
CN105869652A (zh) 心理声学模型计算方法和装置
Luo et al. Audio-visual speech separation using i-vectors
CN111968627B (zh) 一种基于联合字典学习和稀疏表示的骨导语音增强方法
Lezzoum et al. Noise reduction of speech signals using time-varying and multi-band adaptive gain control for smart digital hearing protectors
Ramar et al. A Hybrid MFWT Technique for Denoising Audio Signals
TWI749547B (zh) 應用深度學習的語音增強系統
CN108205127B (zh) 一种基于稀疏表示的水声信号处理方法
Ghamry FPGA Implementation of Hearing Aids using Stationary Wavelet-Packets for Denoising
Kendrick et al. Hearing aid speech enhancement using u-net convolutional neural networks

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 11875486

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 10/10/14 )

122 Ep: pct application non-entry in european phase

Ref document number: 11875486

Country of ref document: EP

Kind code of ref document: A1