Advancement of Two-Stage Binaural Speech Enhancement (TS-BASE) for High-Quality Speech Communication
In this lecture, we introduce our on-going research on two-stage binaural speech enhancement, referred to as TS-BASE, for speech communication. The TS-BASE approach was originally developed from the standpoint of microphone-array signal processing. Though the original TS-BASE algorithm has much evolved recently, all these TS-BASE algorithms consist of the interference estimation through equalizing and canceling the target signal in the first stage and a speech enhancer shared in two channels in the second stage. The equalization and cancellation processes performed for the target signal enable the TS-BASE algorithms independent of the number, location and content of interference signals, and enable to deal with non-stationary multiple interference signals. The shared use of a speech enhancer in two channels enables the TS-BASE algorithms to preserve the binaural cues after processing. Differences between the developed TS-BASE algorithms are mainly in the speech enhancement stage, which includes the techniques using speech absence probability (TS-BASE/SAP), the technique with a soft decision mask filter (TS-BASE/SDBF), the one with a time-variant Wiener filter (TS-BASE/WF). More recently, we revisited our TS-BASE research from the psychoacoustic point of view, and found that the interference estimation stage can be regarded as an application of the “equalization-cancellation (EC) theory that is widely used to explain the “binaural masking level difference (BMLD). At last, the open issues on the TS-BASE algorithms are also extensively discussed in order to show the directions that our further research is going on.
Junfeng Li Shuichi Sakamoto Satoshi Hongo Masato Akagi Y(o)iti Suzukib
School of Information Science,Japan Advanced Institute of Science and Technology Research Institute of Electrical Communication,Tohoku University School of Information Science,Miyagi National College of Technology
国际会议
The 10th Western Pacific Acoustics Conference(第十届西太平洋声学会议WESPAC X)
北京
英文
1-6
2009-09-21（万方平台首次上网日期，不代表论文的发表时间）