The major overhead involved in porting a TTS system from one Indian language to another is in the manual segmentation phase. Manual segmentation of any huge speech corpus is time consuming, tedious and dependent on the person who is segmenting. This correspondence aims at reducing this overhead by automating the segmentation process. A 3-stage, explicit segmentation algorithm is proposed, which uses Quatiere’s sinusoidal model of speech in conjunction with a distance function obtained from Bach scale filter bank, to force align the boundaries of phonemes between 2 stop consonants. Preliminary results for Hindi and Tamil sentences show that the misclassified frames (25ms) per sentence or the Frame Error Rate (FER) is 26.7% and 32.6% respectively.