Echo time delay detection method

Abstract

The invention relates to the field of audio signal processing, in particular to an echo time delay detection method. An audio receiving loudspeaker and a microphone are provided. The echo time delay detection method comprises the following steps: S1, segmentally processing a receiving signal acquired by the audio receiving loudspeaker to obtain a plurality of buffer queues Xi; S2, transforming a transmitting signal to be transmitted into a plurality of processing units Yi according to the same segments by the microphone; S3, implementing similarity calculation on the processing units Yi and the plurality of buffer queues Xi to obtain a maximum value covmax; and S4, performing calculation on each frame of data acquired by the microphone with a Gaussian model to obtain a maximum value f(max)of f(x), and comparing with covmax to obtain an echo time delay estimated value. According to the echo time delay detection method disclosed by the invention, an echo manner is judged comprehensivelyby the Gaussian model and a covariance, so that frequency spectrums with relatively large differences can be judged accurately; voice frequency spectrums with relatively low energy can be judged sensitively; and particularly a more remarkable effect is achieved specific to the echo problem of an interphone.

Description

technical field [0001] The invention relates to the field of audio signal processing, and more specifically, to a method for detecting echo delay. Background technique [0002] Echo is the most common problem in voice calls or voice conferences. Locally generated audio information is transmitted to the remote end through the network, and the remote audio signal is collected by the remote microphone to the remote system through reflection, and then transmitted to the remote system through the IP network. Local, after local playback, it is collected by the local microphone, thus forming a closed-loop positive feedback. When the delay is small, the signal is continuously strengthened, resulting in howling phenomenon. When the delay is large, an echo is generated. [0003] Instant messaging is generally for two-party or even multi-party real-time voice communication. Even if an external speaker is used to play sound in a highly demanding occasion, it is easy to produce an echo p...

AI Technical Summary

Problems solved by technology

However, when the covariance method is used, when the characteristics of the spectrum are greatly different, it cannot be judged accurately, especially for the speech spectrum with low energy.

Images