Structural Feature Parameters Of Speech Signal

The feature parameters of the speech signal as a part of speech signal processingtechnology, plays an absolutely necessary role. A suitable feature parameter can effectivelyshows the the speech signal, and helps people understand more about the information of thespeech in the application of speech visualization.First, based on the study of Chinese phonetics, we find a new feature parameter torecognise the tip of the tongue consonants in Mandarin Chinese-concentrated frequencyarea, and it can distinguish between retroflexes and blade-alveolars. This feature parameteris based on the29-dimensional Mel energy spectrum, shows the concentrated frequencyarea of subband. The result of the test tells that the recognition average rate is about85%,which has potential applications in engineering.Second, for the recognition of Mandarin Chinese vowels, we propose the featureparameter-formant Mel subband. The feature parameters base on the40-dimensional Mellogarithmic energy spectrum, show the Mel subands that formant frequencies are in. Then,compare the formant frequencies and formant Mel subbands, and do researches on itsapplication in vowels visualization and analyse the result.