Segmentation of Malay Syllables in Connected Digit Speech Using Statistical Approach

摘要: This study present segmentation of syllables in Malay connected digit speech. Segmentation was done time domain signal using statistical approaches namely the BrandtA¢â‚¬â„¢s Generalized Likelihood Ratio (GLR) algorithm and Divergence algorithm. These basically detect abrupt changes energy order to determine points. Patterns used this experiment are digits 11 speakers spoken read mode lab environment spontaneous classroom environment. The aim is get close match between reference points automatic Experiments were conducted see effect number auto regressive model p sliding window length L giving better paper reports finding four criterions ie. insertion, omissions, accuracy algorithms. result shows that divergence performed only slightly has opposite testing parameter compared GLR. Read comparison less omission but more insertion.