Noise robust semi-adaptive sound reproduction system based on semi-BSS

This paper proposes a reference signal extraction method for a semi-adaptive sound reproduction system based on semi-blind source separation (semi-BSS) under a noisy environment. Since the fluctuation of room transfer functions degrades the reproduced sound, we have proposed a semi-adaptive sound reproduction system that updates the inverse filters. However, in a noisy environment, it is difficult to observe only the reference signal. Therefore, we introduce semi-BSS based on frequency domain independent component analysis (FDICA) to the semi-adaptive sound reproduction system for extraction of the reproduced signal from the noisy observed signal. First, we obtain the noise signal from the noisy observed signal by semi-BSS. Next, the estimated reference signal is obtained by subtracting the noise signal from the observed signal. In a simulation using real environmental data, the proposed method can extract the reference signal with a high signal-to-deviation ratio (SDR) when RTFs were changed by temperature fluctuation.