Non-air conduction speech reinforcement method based on multi-band spectrum subtraction
An air conduction and speech enhancement technology, applied in speech analysis, instruments, etc., can solve problems such as head movement, gestures, lip movement, heartbeat, breathing, affecting the quality of speech signals, and complex signal components when speaking, to achieve The effect of broad application prospects and market promotion value
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Publication Date
- 2008-12-10
- Estimated Expiration
- Not applicable · inactive patent
Smart Images
Figure 1 Figure 2 Figure 3
Abstract
Description
technical field
[0001] The invention belongs to the technical field of voice signal detection and collection and voice enhancement, in particular to a non-air conduction voice enhancement method based on multi-band spectrum subtraction. Background technique
[0002] Due to the essential difference in the detection principles and methods, the non-air conduction detection technology of voice essentially breaks through the fixed mode and limitations of traditional air conduction voice, and is also an effective supplement to the singleness of traditional detection methods. The non-air conduction voice detection technology can not only overcome the shortcomings and deficiencies of various voice detection modes (it enables people to get rid of the limitation that microphones must be placed near the sound source to detect voices at a longer distance, has strong anti-acoustic interference ability, and overcomes the Laser, infrared, ultrasonic and other detection technologies are vul...
Examples
Embodiment Construction
[0022] The present invention will be described in further detail below in conjunction with the accompanying drawings.
[0023] see figure 1 , the present invention at first will receive the radar non-air conduction voice signal to carry out digital processing and add window and do fast Fourier transform 1; Carry out voice segment and silent segment discrimination to the transformed signal, do multi-band segmentation 2 to the voice segment signal spectrum 3. Do noise estimation 3 to the signal without speech segment and extract the phase information 4 of the speech; do multi-band segmentation to the signal spectrum of the speech segment and use the adjustment coefficient δ of the i-th frequency band i It is weighted by 5; after the noise estimation of the non-speech segment signal, the oversubtraction factor α of the i-th frequency band is used i Weighting it 6; subtracting the weighted noise power spectrum estimated value 7 from the weighted noise power spectrum of the speech...