Persian speech sentence segmentation without speech recognition

In this paper, we propose a method for detection of Persian speech sentence boundaries using a set of prosodic features and spectral centroid. No speech recognizer is used inour proposed method. Silent regions are first detected using four features including spectral centroid, zero crossing rate, energyand pitch. Then, twelve prosodic features are extracted from each silent region. Silent regions may correspond to a sentenceboundary or other regions inside a sentence. Features of Silenceregions of speech data from some speakers are extracted and labeled as silence in the boundary or inside the sentences. Thesefeature vectors and a nonlinear support vector machine (SVM) classifier, is trained and then evaluated for detection of Persianspeech sentence boundaries. The proposed algorithm was evaluated on six speakers from Large FARSDAT data set. Aperformance of 82.4% F-measure was achieved on test set from all speakers in training data and 73.02% F-measure on speakers outside the training data.

کلیدواژه ها:

sentence detection ، prosodic features ، Persian ، SVM classifier