SYSTEM FOR CONVERSION OF MURMURED OR WHISPERED OR BREATHY SPEECH TO NEUTRAL SPEECH AND THEN TO NORMAL / NATURAL SPEECH.

摘要

The system is designed for converting a murmured /whispered/ breathy speech to a normal speech so that it is clearly heard. The system uses murmured speech as input. The system consists of number of blocks namely removing the spikes, isolation of unvoiced segments from voiced and silence, isolation of voiced vowels and low amplitude voiced consonants from silence, removing the clicks at the start of the voiced part and giving artificial voicing to the voiced sections. Before the artificial voicing, the fundamental frequency of the original signal if available is tracked or use normal pitch frequency for a male or a female and is then used for giving the excitation. This excitation is used with modified Linear Predictor Coefficients (LPCs) and modified gain contour to synthesize voiced sections. This generates the neutral speech. It can be further converted to normal by changing the pitch contours of the start and end words and the in between words to give normal speech effect. The system generates a speech closer to the original normal speech if the properties of the converted vowels and the unvoiced sections are changed further. The system can be further tuned as per the gender of the speaker.