| 摘要 |
Invention refers to a speaker localization system and procedure in open or closed room, which is based on the grounds of implementation of a set of microphone in sequence with arbitrary number of microphones and which may be applied in control and management of robots, video surveillance and processes that demand interactive information on localization of a speeker, or in ‘hands-free’ communication systems such as teleconference systems, video conference systems, speakerphones, and alike. The system consists of a microphone in series, activation module and a module for treatment of acoustic and speech signals. The procedure is based on the cross correllation, i.e. phase transformation PHAT, optimized according to characteristics of speech signal and on a detector of speech activity (VAD) on the base of superdirective rectifier (SD-BF). Optimization of PHAT cross correlation achieves improved accuracy and more precise estimation of angle and azimuth &thetas; while application of SD-BF in VAD improves the reliability of estimation in dynamic conditions of application of the system for localization of a speaker. Determination of a filtration function W(n) Pretreatment of a signal and FFT Cross correlation analysis → PHAT → Azimuth estimation → VAD
|
