Method and device for voice activity detection (VAD) and encoder

Abstract

The embodiment of the invention discloses a method and device for voice activity detection (VAD) and an encoder. The method for VAD comprises the following steps: when an input signal is a background noise, acquiring a fluctuation characteristic value of the background noise, wherein the fluctuation characteristic value is used for indicating the fluctuation amplitude of the background noise; carrying out self-adaptive regulation on relevant parameters of a determining criterion of VAD in accordance with the fluctuation characteristic value; and carrying out VAD determination on the input signal based on the relevant parameters of the determining criterion subject to self-adaptive regulation. The embodiment of the invention can be self-adapted to the fluctuation of the background noise so as to carry out VAD determination, thereby improving the performance of VAD determination, saving the limited channel bandwidth resources and realizing the high-efficiency utilization of the channel bandwidth.

Description

technical field [0001] Embodiments of the present invention relate to communication technology, in particular to a voice activation detection method and device, and an encoder. Background technique [0002] In a communication system, especially a wireless communication system or a mobile communication system, channel bandwidth is a scarce resource. According to statistics, in a two-way conversation, the speaking time of both parties accounts for only about half of the total conversation time, and the other half is in a silent state. Since the communication system only transmits signals when people are talking, and stops signal transmission when there is silence, it cannot allocate the bandwidth occupied by silence to other communication services, seriously wasting limited channel bandwidth resources. [0003] In order to make full use of channel resources, in the current technology, voice activity detection (VAD) technology is used to detect when the two parties in the call...

AI Technical Summary

Problems solved by technology

[0005] But, in the process of realizing the present invention, the inventor finds, because different types of background noises have different fluctuations, that is: the magnitude of fluctuations, even at the same level, different types of background noises will affect the decision efficiency of VAD. have a greater impact on the accuracy of the decision

However, when the existing AMR performs VAD judgment, it can only adapt to the level of background noise, but cannot adapt to the fluctuation of background noise. Therefore, the performance of VAD judgment on input signals with different types of background noise is quite different.

For example: under the same level of background noise, AMR has a very high VAD decision performance when the background noise is car noise, but when the background noise is babble noise, the VAD decision performance is greatly reduced, Caused a huge waste of channel bandwidth resources

Images