Quick and stable echo eliminating method of intelligent voice interaction equipment

Abstract

The invention discloses a quick and stable echo eliminating method of intelligent voice interaction equipment. The method comprises the steps of playing a preset voice content by the intelligent voiceinteraction equipment after first starting; suppressing self noise of the played preset voice content after first starting by the intelligent voice interaction equipment through adaptive filtering technology, when the self noise reaches a preset attenuation value, storing a filter coefficient at this time point, and obtaining a stored filter coefficient; in using the intelligent voice interactionequipment, using the stored filter coefficient as an initial coefficient of the adaptive filtering technology; and furthermore in a tracking process, limiting the increment of the filter coefficientto a preset range. According to the quick and stable echo eliminating method, through setting the specific initial coefficient of a filter, convergence speed of the adaptive filter can be increased; and furthermore in the tracking process, the increment range of the filter coefficient is limited, thereby preventing divergence of the adaptive filter.

Description

technical field [0001] The invention relates to the technical field of voice signal processing, in particular to a fast and stable echo cancellation method for an intelligent voice interactive device. Background technique [0002] Smart voice devices usually communicate with people through voice, and the sound from the smart voice device will be transmitted to the system through the microphone. In order to avoid recognizing it as the user's voice, it needs to be removed by echo cancellation technology. If the removal is not clean, the voice recognition system will recognize the voice of the device itself, resulting in confusion in the interaction. The self-pronunciation of intelligent voice equipment is usually processed by echo cancellation technology. Its core technology includes adaptive filtering technology, and adaptive filtering technology takes a period of time to converge, and when there is noise other than its own voice mixed in, it will cause the divergence of the ...

AI Technical Summary

Problems solved by technology

If the removal is not clean, the voice recognition system will recognize the voice of the device itself, resulting in confusion in the interaction

The self-pronunciation of intelligent voice equipment is usually processed by echo cancellation technology. Its core technology includes adaptive filtering technology, and adaptive filtering technology takes a period of time to converge, and when there is noise other than its own voice mixed in, it will cause the divergence of the filter. , leading to more residual noise

[0003] The current echo cancellation technology has a slow convergence speed in the initial stage, resulting in more residual self-noise, which is likely to cause self-excitation of the speech recognition system and cause interaction confusion. It is easy to cause the filter to diverge, resulting in more self-noise residuals

Images