Research On The Algorithms Of Passive Acoustic Localization In Rooms

Acoustic localization in rooms is a new technology, many new methods now have been developed, such as Higher-Order Spectral Analysis method, adaptive-filter delay-estimation method, and high resolution spectral estimation. Although beam-forming method is not limited by correlation, but it will increase the noise while it increase the useful signal, as well as it will be limited by the base array aperture; Adaptive-filter delay-estimation method could resolve the real-time issue for locomotive acoustic source, but it's based on noise non-correlation and it's constringency speed is slow; High resolution spectral estimation has complex algorithms and it's performance will be rapidly degraded when signals are nonstationarity.In order to make the acoustic localization system in rooms better, a bi-sensor array time-delay estimation based on wavelet transform is designed in this thesis. It is good for delay-estimation of signal from cross-correlation noise. But this method has lower performance due to the limitation of wavelet analysis itself (wavelet is chosen according to experience, and it is non-adaptive).After studying on many algorithms, a new method named time-delay localization based on Hilbert-huang transform is designed in this thesis, which is mainly used to overcome the nonstationarity of speech signals. This method works through performing a time adaptive decomposition operation named empirical mode decomposition (EMD) on the signals; Then the signals will be decomposed into a set of complete and almost orthogonal components named intrinsic mode function (IMF). The IMF signals will later be de-noised by wavelet soft threshold method to improve accuracy. In this way, the time-delay estimation can be calculated through cross-correlation principle. The acoustic azimuth angle is last got through geometry interpolation. The system is adaptive because of IMF. In addition, the details of signals can be illustrated accurately in Hilbert edge spectrum.A bi-sensor array mode is established in this thesis. On the foundation, the new method named time-delay localization based on HHT, has been implemented, which has sharp correlation peak, and it's localization error is 0.14%.