This paper addresses the problem of binaural speech sound source localization (SSL) in noisy and reverberant environments. For the binaural setup, the array response corresponding to the direct-path sound propagation of a single source is a function of the source direction. In practice, this response is contaminated by noise and reverberation. The direct-path relative transfer function (DP-RTF) is defined as the ratio between the direct-path acoustic transfer function (ATF) of the two channels, and it is an important feature for SSL. We propose a method to estimate the DP-RTF from the noisy and reverberant sensor signals in the short time Fourier transform (STFT) domain. First, the convolutive transfer function (CTF) approximation is adopted to accurately represent the impulse response of the sensor array in the STFT domain. The first element of the CTF is mainly composed of the direct-path ATF. The DP-RTF is then estimated by using the auto and cross power spectral density (PSD) of multiple STFT frames at each frequency. In the presence of stationary noise, an inter-frame spectral subtraction algorithm is proposed, which enables to achieve the estimation of noise-free auto and cross PSD. Finally, the estimated DP-RTFs are concatenated across frequency and used as a feature vector for SSL. Experiments show that the resulting SSL method performs well even under severe adverse acoustic condition, and outperforms the comparison methods under most of the acoustic conditions.
View on arXiv