Abstract—The most important part in speech applications is parameter extraction, but the existence of impulsive noise can influence the parameter extraction dramatically. To minimize the influence caused by the impulsive noise, this paper proposed a new impulsive noise detection and location method that incorporates the advantages of both model-based and statistics-based methods. First the Projection Statistics (PS) is calculated to detect and locate the impulsive noise as outliers from statistics view, and then the residual signals of speech is calculated and compared with a pre-defined threshold to refine the raw result. Experiments show that the proposed method is obviously better than the model-based and statistics-based methods in detection accuracy and location accuracy, especially in the false alarm rate.

 

Index Terms—speech applications, impulsive noise, Projection Statistics, residual signals

Cite: Liang Chang, Kun Tang, and Huijuan Cui, "Impulsive Noise Detection and Location in Speech," International Journal of Signal Processing Systems, Vol. 4, No. 6, pp. 446-450, December 2016. doi: 10.18178/ijsps.4.6.446-450