Speech enhanced interaction method, system, storage medium and electronic device

Abstract

The invention provides a speech enhanced interaction method, a speech enhanced interaction system, a storage medium and an electronic device. The method includes the following steps that: the time-domain signals of microphones in an annular microphone array are converted into the frequency-domain signals of the microphones, and reverberation suppression and stationary noise suppression are performed on the frequency-domain signals of the microphones; wake-up direction sound source positioning is performed based on the reverberation and stationary noise-removed frequency-domain signals of the microphones, so that a wake-up direction is obtained; main direction beam time-domain signals and wake-up direction beam time-domain signals are obtained in a main direction and the wake-up direction on the basis of the reverberation and stationary noise-removed frequency-domain signals of the microphones; speech recognition is performed on the main direction beam time-domain signals; and wake-up word recognition is performed on the wake-up direction beam time-domain signals, and if the signals are identified as wake-up words, the main direction is changed to the obtained wake-up direction. With the speech enhanced interaction method, the speech enhanced interaction system, the storage medium and the electronic device of the invention adopted, the stability and reliability of speech interaction can be improved effectively.

Description

technical field [0001] The invention relates to the technical field of speech processing, in particular to a speech enhancement interaction method and system, a storage medium and electronic equipment. Background technique [0002] With the development of information technology, artificial intelligence technology has increasingly entered people's lives. Among the many human-computer interactions, voice interaction is the most natural and most in line with human behavior. The continuous development of voice recognition technology has also made voice interaction a reality. During use, a specific wake-up word is usually used to trigger the voice interaction system. However, in real life scenarios, the voice interaction environment is relatively complex and is easily affected by environmental noise, reverberation, and human voice interference, which makes the signal-to-noise ratio of the voice signal collected by the microphone poor, which seriously affects the accuracy of voi...

AI Technical Summary

Problems solved by technology

[0004] The single-microphone voice enhancement technology has a small algorithm complexity, but the suppression effect on non-stationary noise is poor, and it is easy to cause different degrees of voice distortion

Due to the limitations of the linear microphone array speech enhancement method, the effect of picking up sound in the normal direction of the array is better, but the effect of picking up sound in the axial direction is poor; at the same time, although adaptive beamforming has real-time tracking of noise, However, when the noise environment is complex and the reverberation is large, it is difficult to guarantee the accuracy of sound source localization, and the accuracy of adaptive tracking in the null direction is also difficult to guarantee, resulting in different degrees of voice distortion.

In order to avoid the deterioration of the effect of adaptive beams in complex environments, it is usually necessary to design complex adaptive algorithms and complex sound source localization methods, which are difficult to meet the real-time application requirements of embedded systems

In general, the sound source localization method and the adaptive beam method usually use different design methods, such as the GCC and GSC methods, and there are almost no multiplexing modules in the two, resulting in greater complexity

Images