Abstract
The quality and intelligibility of the speech are usually impaired by the interference of background noise when using internet voice calls. To solve this problem in the context of wearable smart devices, this paper introduces a dual-microphone, bone-conduction (BC) sensor assisted beamformer and a simple recurrent unit (SRU)-based neural network postfilter for real-time speech enhancement. Assisted by the BC sensor, which is insensitive to the environmental noise compared to the regular air-conduction (AC) microphone, the accurate voice activity detection (VAD) can be obtained from the BC signal and incorporated into the adaptive noise canceller (ANC) and adaptive block matrix (ABM). The SRU-based postfilter consists of a recurrent neural network with a small number of parameters, which improves the computational efficiency. The sub-band signal processing is designed to compress the input features of the neural network, and the scale-invariant signal-to-distortion ratio (SI-SDR) is developed as the loss function to minimize the distortion of the desired speech signal. Experimental results demonstrate that the proposed real-time speech enhancement system provides significant speech sound quality and intelligibility improvements for all noise types and levels when compared with the AC-only beamformer with a postfiltering algorithm.
Highlights
In recent years, the signal transmission bandwidth and network technology have been significantly improved, and communication system can real-time transmit speech signals with a higher sampling rate and deeper sampling bit depth
In the non-stationary noiseTo sound field,a real-time simple recurrent unit (SRU)-based neural network noise reduction algorithm with a small when the noise is a non-point source or the noise source is in the same direction with the desired signal, number parameters, this paper the obtaining array postfiltering as athe single-channel noise provided reduction the array of postfiltering algorithm has treats difficulty benefit from spatial information task
It can be seen that the deep neural network (DNN)-based method is significantly better than the traditional optimally modified log-spectral amplitude (OMLSA) method, because the conventional method depends on the assumption that noise changes much more slowly than speech
Summary
The signal transmission bandwidth and network technology have been significantly improved, and communication system can real-time transmit speech signals with a higher sampling rate and deeper sampling bit depth. In the non-stationary noiseTo sound field,a real-time simple recurrent unit (SRU)-based neural network noise reduction algorithm with a small when the noise is a non-point source or the noise source is in the same direction with the desired signal, number parameters, this paper the obtaining array postfiltering as athe single-channel noise provided reduction the array of postfiltering algorithm has treats difficulty benefit from spatial information task. With the assistance single-channel noise reduction system based on SRU was used as a postfilter to eliminate the residual of the BC signal, the noise suppression capability of the dual-microphone GSC system under low noise in the microphone array output.
Sign-in/Register to view highlights & summary 
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call