In noisy and reverberant environment, to improve the performance of speech enhancement algorithm, multi-channel system has been proved effective. Classical beamforming techniques are often available ways to tackle the noise. But beamforming alone cannot supply sufficient noise reduction, when it is presented with non-directional noise such as diffuse noise, reverberation and residual directional noise. Thus, post-filtering is often required.
Recently, researchers from Institute of Acoustics of the Chinese Academy of Sciences have introduced an improved multi-channel post-filtering algorithm adapting to different reverberant environments for distant-talking scene (both reverberation and noise are existing). The improved algorithm can offer robust speech enhancement in noisy and reverberant environment.
In order to accurately obtain the estimation of the highly non-stationary and residual directional noise, signal presence probability, which is often seen as a controller, is deduced according to reasonable signal detection scheme.
Since the spatial filtering response of each frequency bin is different, threshold determining the tradeoff between false alarm and detection probability should be adaptively confirmed. It is rather than using empirical preset parameter of the whole frequency band.
Furthermore, direct-to-reverberate ratio is introduced into the signal detection scheme so that the proposed algorithm is more robust to reverberation. The reason for that is reflections would have reduced the discrimination between desired speech and competing interference.
Essentially, the procedure of obtaining the modified signal presence probability can be shown by Fig.1. The a priori signal absence probability is obtained using the soft-decision information of both transient beam-to-reference ratio and local non-stationarity through signal detection procedure, and further the signal presence probability is derived. Since the key detection thresholds for estimating the presence of desired signal from look direction are modified combining direct-to-reverberate ratio which is blindly estimated, a reverberation robust multi-channel post-filtering technique is achieved based on the modified signal presence probability.
Fig.1 Diagram of the derivation of modified signal presence probability (Image by WANG)
In the experiments, the microphone array used in this research is composed of 4 omnidirectional MEMS microphones in broadside orientation. The distance between the microphones is set to be 5cm. The experiments take place in an office room of 6m × 5m × 3m.
Two cased are taken into consideration. In the first case, two interferences (a competing speaker and a gauss white noise source) are located in 90◦ and 45◦ of the array respectively with reverberation time of 300ms and 600ms (T60), where the reverberation time is controlled with sound absorption material. In the other case, the two interferences are located in 30◦ and 20◦, since the spatial distinction reduces due to the small angle between target signal and interference.
Experimental results show that the proposed multi-channel speech enhancement algorithm provides a robust performance in different reverberant environment with directional interference and stationary diffuse noise in terms of the speech quality measures (segmental SNR, noise reduction and log-spectral distance).
Reference:
WANG Xiaofei, GUO Yanmeng, Fu Qiang, YAN Yonghong. Speech Enhancement Using Multi-channel Post-Filtering with Modified Signal Presence Probability in Reverberant Environment. Chinese Journal of Electronics (Vol. 25, No. 3, May 2016, pp. 512-519). DOI: 10.1049/cje.2016.05.017
Contact:
WANG Xiaofei
Institute of Acoustics, Chinese Academy of Sciences, 100190 Beijing, China
Email: wangxiaofei@hccl.ioa.ac.cn