ADDINGAN EMOTIONS FILTERTO MALAY TEXT-TO-SPEECH
Zuraidah Mohd Don, Gerry Knowles · 2007
Inthis paperwepresent thefindings ofourresearch whichaimstodevelop anemotions filter that canbe addedtoanexisting MalayText-to-Speech system to produce anoutput expressing happiness, anger, sadness andfear. Theendgoal istoproduce anoutput that isas natural aspossible, thuscontributing towards the enhancement oftheexisting system. Theemotions filter wasdeveloped bymanipulating pitch andduration ofthe output using arule-based approach. Thedata wasmade upofemotional sentences produced byafemale native speaker ofMalay. Theinformation extracted fromthe analysis wasusedtodevelop thefilter. Theemotional speech output underwent several acceptance tests. The results showedthattheemotions filter developed was compatible withFASIHandother TTSsystems using the rule-based approach ofprosodic manipulation. However, further workneedtobedonetoenhance thenaturalness oftheoutput. IndexTerms-Text-to-Speech, MBROLA, emotionsFilter, prosody, rule-based