Improved Frequency Domain Blind Source Separation for Audio Signals via Direction of Arrival Knowledge
Ricardo Kehrle Miranda, Soyuj Kumar Sahoo, Ricardo Zelenovsky, João Luís Rodrigues Paulo · 2013
The sound pollution is a common and se- rious problem in modern cities. Examples of sources of sound pollution are other speakers, traffic and household devices such as vacuum cleaner, air conditioner, TV and radios. These sound interference sources degrade the per- formance of audio devices such as hearing aids, smart TVs and forensic recorders. One way to tackle this prob- lem is via Blind Source Separation (BSS) algorithms us- ing microphone arrays. In this paper, we propose a tech- nique to improve the performance of frequency domain BSS (FDBSS) by taking into account the direction of ar- rival (DOA) knowledge. With our proposed scheme, a sig- nificant improvement is obtained for end fire angles.