A Probabilistic Model for Sound Direction of Arrival Estimation Based on Signal-to-noise Ratios in the Frequency Domain

• Main Authors: Li, Chao-Wen, 李詔文
• Other Authors: Liu, Yi-Wen
• Format: Others
• Language: en_US
• Published: 2015
• Online Access: http://ndltd.ncl.edu.tw/handle/17011033605974216183

碩士 === 國立清華大學 === 電機工程學系 === 103 === This thesis proposes a method that estimates the time difference of arrival (TDOA) using the interaural phase difference as a cue to locate the direction of a sound source. The algorithm fist estimates the amplitude of the Fourier transform of the signal to obtain an approximated signal to noise ratio (SNR). The phase difference of each frequency bin is then modeled in terms of its probability density function. Through combining the information from all frequency bins, the probability density function of TDOA can be obtained. Instead of giving a direction as the final result, the probability density function provides a measure of confident in estimation. The Gaussian function is used to approx- imate the probability density function of phase in order to simplify the computation. This makes it possible to implement the algorithm in real-time. The experiment result also shows that the algorithm is more robust to noise in comparison with an existing algorithm published by [1], which uses power weighting and interaural phase spectrum (IPD) to determine interaural time difference (ITD).