A novel spectral subtraction scheme for robust speech recognition: spectral subtraction using spectral harmonics of speech

The weakness of conventional spectral subtractive-type algorithm is identified and presented in Section 2. The proposed remedial approach is described in Section 3. In Section 4, we show the proposed method’s effectiveness over conventional methods with representative experiments using Aurora 2. Concluding remarks are provided in Section 5. This paper addresses a novel noise-compensation scheme to solve the mismatch problem between training condition and testing condition for the automatic speech recognition (ASR) system, specifically in the car environments. The conventional spectral subtraction schemes rely on the signal to noise ratio (SNR) such that attenuation is imposed on that part of the spectrum that appears to have low SNR, and accentuation is made on that part of high SNR. However, since these schemes are based on the postulation that the power spectrum of noise is in general at the lower level in magnitude than that of speech. Therefore, while such postulation is adequate for high SNR environment, it is grossly inadequate for low SNR scenarios such as car environment. This paper proposes an efficient spectral subtraction scheme focused to specifically low SNR noisy environments by distinguishing the speech-dominant segment from the noisedominant segment in speech spectrum. Representative experiments confirm the superior performance of the proposed method over conventional methods. The experiments are conducted using car noise-corrupted utterances of Aurora2 corpus. 2. SPECTRAL SUBTRACTIVE-TYPE ALGORITHM When speech x is corrupted by background additive noise , the corrupted speech can be expressed as follows: ) (n ) (n b ) ( ) ( ) ( n b n x n y + = (1) If speech and noise are assumed to be uncorrelated, in frequency domain, it can be represented as follows: 2 2 2 | ) ( | | ) ( | | ) ( | k B k X k Y + = (2) where is index of frequency bin. k 2.1. Spectral Subtraction