8th International Conference on Spoken Language Processing

Jeju Island, Korea
October 4-8, 2004

A Noise-Robust Feature Extraction Method Based on Pitch-Synchronous ZCPA for ASR

Ghulam Muhammad, Takashi Fukuda, Junsei Horikawa, Tsuneo Nitta

Toyohashi University of Technology, Japan

In this paper, we propose a novel feature extraction method based on an auditory nervous system for robust automatic speech recognition (ASR). In the proposed method, a pitch-synchronous mechanism is embedded in ZCPA (Zero-Crossings Peak-Amplitudes), which has previously been shown to outperform the conventional features in the presence of noise. A noise-robust non-delayed pitch determination algorithm (PDA) is also developed. In the experiment, the proposed pitch-synchronous ZCPA (PS-ZCPA)was proved more robust than the original ZCPA method. Moreover, a simple noise subtraction (NS) method is also integrated in the proposed method and the performance was evaluated using the Aurora-2J database. The experimental results showed the superiority of the proposed PSZCPA method with NS over the PS-ZCPA method without NS.

Full Paper

Bibliographic reference.  Muhammad, Ghulam / Fukuda, Takashi / Horikawa, Junsei / Nitta, Tsuneo (2004): "A noise-robust feature extraction method based on pitch-synchronous ZCPA for ASR", In INTERSPEECH-2004, 133-136.