Low signal to noise ratio voice endpoint detection method based on time-frequency instaneous energy spectrum

Abstract

The invention relates to a low signal to noise ratio voice endpoint detection method based on a time-frequency instantaneous energy spectrum. A traditional method used for reducing noise is not ideal in noise-reducing effect. The detection method comprises that noise-contained voice is resolved firstly and reconstructs a low frequency component which is resolved, by regarding the time-frequency instantaneous energy spectrum as a basis of the endpoint detection. After being reconstructed, the signal is dealt by an empirical mode decomposition and the screening and rejecting of an intrinsic mode function are carried out in order to gain an effective international monetary fund (IMF) component by applying a correlation coefficient threshold value principle. The IMF component gained through reconstruction undergoes a Hilbert transformation and a corresponding instantaneous energy spectrum is calculated to form a feature vector of voice endpoint detection. The result is dealt with the subframe. A former five frames instantaneous energy spectrum mean value is regarded as the instantaneous energy spectrum of the noise and a start-stop dual threshold method is adopted to detect the endpoint of the noise. The detection method has the advantages of improving the noise-contained time-frequency resolving power and improving the efficiency of the voice endpoint detection.

Description

technical field

[0001] The invention belongs to the field of voice processing, and relates to a low signal-to-noise ratio voice endpoint detection method based on time-frequency instantaneous energy spectrum. Background technique

[0002] Various noises will inevitably be introduced in the process of voice acquisition, transmission and communication, and the existence of noise will directly affect the clarity and intelligibility of voice. Endpoint detection of noisy speech signals to obtain the start and end points of effective speech segments plays a very important role in subsequent speech enhancement, coding and recognition. At present, the traditional endpoint detection methods mainly include average energy, average zero-crossing rate, cepstral coefficient, short-time frequency band variance, short-time energy-frequency value, cepstrum distance, autocorrelation similarity distance, information entropy and spectral entropy. But they are all based on the assumption that t...

Examples

Embodiment Construction

[0021] Below in conjunction with accompanying drawing, the present invention will be further described, and the concrete steps of the inventive method are:

[0022] Step (1) For the noisy speech signal under the background of strong noise (Such as figure 1 shown) plus Hamming window processing. Use the db3 wavelet basis function in Daubechies to perform three-layer wavelet packet decomposition on the windowed noisy speech signal, and the schematic diagram of the wavelet packet decomposition binary tree is as follows figure 2 shown. Reconstruct the decomposed result to obtain the reconstruction signal, denoted as , and the corresponding frequency bands are ,in for the minimum frequency resolution, , is the sampling frequency.

[0023] Step (2) will reconstruct the obtained low frequency component signal Perform adaptive EMD decomposition (the first 7 IMF components such as image 3 shown), thus obtaining a finite number of IMF components and residual signal ...

technical field

[0001] The invention belongs to the field of voice processing, and relates to a low signal-to-noise ratio voice endpoint detection method based on time-frequency instantaneous energy spectrum. Background technique

[0002] Various noises will inevitably be introduced in the process of voice acquisition, transmission and communication, and the existence of noise will directly affect the clarity and intelligibility of voice. Endpoint detection of noisy speech signals to obtain the start and end points of effective speech segments plays a very important role in subsequent speech enhancement, coding and recognition. At present, the traditional endpoint detection methods mainly include average energy, average zero-crossing rate, cepstral coefficient, short-time frequency band variance, short-time energy-frequency value, cepstrum distance, autocorrelation similarity distance, information entropy and spectral entropy. But they are all based on the assumption that t...

Examples

Embodiment Construction

[0021] Below in conjunction with accompanying drawing, the present invention will be further described, and the concrete steps of the inventive method are:

[0022] Step (1) For the noisy speech signal under the background of strong noise (Such as figure 1 shown) plus Hamming window processing. Use the db3 wavelet basis function in Daubechies to perform three-layer wavelet packet decomposition on the windowed noisy speech signal, and the schematic diagram of the wavelet packet decomposition binary tree is as follows figure 2 shown. Reconstruct the decomposed result to obtain the reconstruction signal, denoted as , and the corresponding frequency bands are ,in for the minimum frequency resolution, , is the sampling frequency.

[0023] Step (2) will reconstruct the obtained low frequency component signal Perform adaptive EMD decomposition (the first 7 IMF components such as image 3 shown), thus obtaining a finite number of IMF components and residual signal ...

Images