A Speech Enhancement System for Distributed Wireless Acoustic Sensor Networks

Abstract

The invention discloses a speech enhancement system for a distributed wireless acoustic sensor network, which specifically includes: a phase alignment module, an analytic transformation module, a wavelet transformation module, an NPSD estimation module, an MVDR filter module, a distributed algorithm iteration module and a wavelet Inverse transform module. Using this system, the input signal of each node must be phase-aligned with the node farthest from the sound source, and each signal after alignment is subjected to analytic wavelet transformation; then, the NPSD is estimated in the wavelet domain, and the MVDR filter coefficient is obtained at the same time ; Finally, the output signal of each node microphone is obtained through the distributed algorithm iteration module, and the output signal is inversely transformed by wavelet and the real part is taken to obtain the time domain signal. The system is implemented by exchanging specific data between each microphone and its nearby microphones, so that the initial state values ​​of all microphones converge to the global average value through iteration.

Description

technical field [0001] The invention relates to the technical field of speech signal processing, in particular to a speech enhancement system used in a distributed wireless acoustic sensor network. Background technique [0002] In practical applications, audio signals received by audio processing devices are often interfered by various noises. Interfering noise will degrade the performance of the device, and in severe cases, the device may not even work properly. In order to overcome the influence of noise, speech enhancement technology has been proposed. Due to the limited performance of speech enhancement methods based on a single microphone, microphone array speech enhancement methods that can utilize spatial information have gained attention. However, the microphone array also exposes some limitations such as the need to know the geometric information of the array and to have a fixed and regular array structure. With the rapid development of wireless sensor technology...

AI Technical Summary

Problems solved by technology

But the disadvantage of this method is that in a rapidly changing noise field, the beamforming output needs to be calculated in blocks

The algorithm is the MVDR beamforming studied in the uniform linear array model, and the application in the distributed algorithm does not have low complexity characteristics

Images