Automatic segmentation and labeling of speech recorded in unknown noisy channel environments

摘要: This paper investigates the problem of automatic segmentation of speech recorded in noisy channel corrupted environments. Several speech enhancement and channel compensation techniques previously proposed for robust speech recognition are evaluated and compared for improved segmentation in colored noise. Speech enhancement algorithms considered include: Generalized Spectral Subtraction, Nonlinear Spectral Subtraction, Ephraim-Malah MMSE enhancement, and AutoLSP Constrained Iterative Wiener ltering. In addition, the coupling of front-end processing such as cepstral mean subtraction in tandem with speech enhancement methods is considered for noisy channel compensation. Compensation performance is assessed for each method using automatic segmentation of the telephone transmitted NTIMIT and cellular telephone transmitted CTIMIT databases.