Spectral envelope silence detection method

Abstract

The invention discloses a spectral envelope silence detection method. The method comprises the following steps: converting a time domain signal into a frequency domain signal and calculating the spectral energy distribution of the signal; performing gain oscillation detection on a received signal according to the state of the received signal, and updating the spectral energy distribution of a background noise by selecting a corresponding updating mode according to the current state; calculating the signal-to-noise ratio of a frame by utilizing the spectral energy distribution of the received signal and the spectral energy distribution of the background noise, and storing the signal-to-noise ratio of the frame within a certain recent period of time; performing spectral envelope multi-shift according to the records of the signal-to-noise ratio of the frame and the signal-to-noise ratio of spectral envelope, and judging an input signal is a voice or a noise according to the state output of the state shift. By the method, the starting and the ending of a voice signal can be detected quickly, and basis can be provided for processing voice signals such as voice noise reduction and voice silence.

Description

technical field [0001] The invention relates to the field of voice signal processing of radio communication, in particular to a spectrum envelope mute detection method. Background technique [0002] Silence detection is based on the characteristic analysis of the signal to distinguish speech signals from non-speech signals. It has a very wide range of applications in the fields of speech coding, speech enhancement, speech recognition and wireless communication. For example, in speech coding, silence detection technology is used to Distinguish between speech and non-speech areas, and use a small number of bits to encode the non-speech areas to ensure that the average transmission rate of speech coding is lower under the premise of the same quality. In speech enhancement, the presence of speech is identified through silence detection, silence and noise statistics are performed on non-speech segments, and speech noise reduction processing is performed on speech segments. In th...

AI Technical Summary

Problems solved by technology

The traditional silence detection method detects by means of short-term energy, zero-crossing rate and correlation. It has a good detection probability under high signal-to-noise ratio, but the detection is not accurate enough under low signal-to-noise ratio, and the noise is judged as speech or Speech judgment into noise will seriously affect the effect of speech coding, enhancement and recognition

Images