Detection of aspiration in rabha alveolar fricatives using zero frequency filtering

Abstract

Even though aspiration occurs commonly with the stop sounds, its occurrence with fricatives is rarely observed. In Rabha language, we observe the occurrence of aspiration as a secondary feature to the voiceless alveolar fricatives. However, distinguishing between an aspirated and unaspirated voiceless alveolar fricative is a difficult task for an untrained ear. Considering the acoustic cues identified for aspirated voiceless alveolar fricatives, a method for automatically detecting aspiration in voiceless alveolar fricatives is proposed in this work. The proposed method, involving zero frequency filtering, exploits the impulse-like excitation in speech and substantiates the concentration of the spectral energy localized in the low-frequency region exhibited by the acoustic phonetic study. The features, viz. amplitude, strength of excitation and variance of the epoch intervals are investigated to detect the presence of aspiration in aspirated fricatives of the Rabha language. Finally, a temporal method is designed to segment the consonant and the vowel regions from the speech segments automatically and to distinguish between the aspirated and the unaspirated groups of fricatives in Rabha.