|
For the doctor of traditional Chinese medicine, four major diagnostic methods, inspection, auscultation, questioning and palpation, are the perfect ways to getpathology. Nevertheless, the skills of auscultation are usually to be omitted by youngdoctors and there are few scholastic papers about auscultation with digital processing of speech signals. Accordingly, in this thesis, we used digital processing techniques of speech signals to give scientific evidences for auscultation. The purpose of this thesis was to find out the speech differences between theupper GI disorder patients and the healthy people. In order to search the speechcharacteristic features about them, we used speech parameters, including pitch, zerocrossing rates and formants, to compare the speech pattern of healthy people with upperGI disorder patients. The vowel (/a/, /i/, /u/, /e/, /o/) sounds were chosen for two subjectgroups to pronounce for 5 minute. There were 52 healthy people and 47 patients withthe upper GI disorder involved in this research. Experimental results showed that:(1)The pitches of healthy males were lower thanthe upper GI disorder patients'' and the pitches of healthy females were higher(P<0.05).(2)The zero crossing rates for vowel (/a/, /i/, /o/) were significantly different between the upper GI disorder patients and healthy people. In addition, the zerocrossing rates were able to distinguish between sex (P<0.001).(3)The first formant forvowel (/a/, /i/, /o/) and the second formant for vowel (/u/) of healthy people were lowerthan the upper GI disorder patients''(P<0.05). For this thesis, why the experimental results showed significant differences? The reasons for the significant differences were the upper GI disorder patients with sharppain on abdomen, cold sweat, accelerated heartthrob, oppression, egesta with blood, andthirsty made correlations between exhalation, the excitation of vocal cords and qivacuity, qi solidity. The statistical results showed that vowel (/o/) were the most representative to distinguish between the diseased and the healthy subjects. The pitch, zero crossing ratesand the first formant of vowel (/o/) of males were used to carry out Back-propagation algorithm of neural network training by PRW (Pattern Recognition Workbench) andreached an accuracy rate above 90% to distinguish between two subject groups. After the processes of speech analysis, statistics and recognition, we made aconclusion that pitch, zero crossing rates, and the first formant of vowel (/o/) were theindices to distinguish between the diseased and the healthy. The pitches of gastritis werelower than the healthy peoples'', but the zero crossing rates and the first formant werehigher than the healthy peoples''. The pitches of some peptic ulcers who are pyrexiogenic were higher than the healthy peoples''; the zero crossing rates and the first formant were also higher than the healthy peoples''.
|