초록

주관성이 개입되지 않으면서 정확성과 간편성이 높은 음성 분석을 통해 사상체질별 모집단을 개발하고, 이 모집단을 바탕으로 사상체질별 알고리즘을 확립해서 객관적인 진단기기를 통해 사상체질을 진단하는 기법을 개발한다.

Abstract ▼

1. The Sasang constitution classification and the Evolution of Population. Through the trust degree, inspection of the Sasang constitution classification inspect paper(QSCC II ) discriminated constitution of the studying suspect that is the regular course 2nd, 3rd, 4th student in the Oriental Medica

1. The Sasang constitution classification and the Evolution of Population. Through the trust degree, inspection of the Sasang constitution classification inspect paper(QSCC II ) discriminated constitution of the studying suspect that is the regular course 2nd, 3rd, 4th student in the Oriental Medical college, the Sangji University and late learners.In view of the result of conviction of correlation between the B.M.I and Sasang constitution. There is significant difference below P<0.001 between Taeumin and Soeumin, Taeumin and Soyangin.The height and the weight effect to the fundamental frequency and the resonate frequency.At the analyzing sound(voice), in according to each constitution, there is a difference from the amplitude of vibration. 2. The analyzing sound of the each Sasang constitutional population. According to the examination that is based on Oriental Medical literature, there is recognized that each constitution manifest the sound characteristic. Each constitution as a significant different point as the average composition reading time, weight, formant frequency l, Bandwidth 2, formant frequency 5, Pitch -Maximum, Pitch Maximum-Pitch Minimum, Energy Mean, etc.By virtue of the constitutional discriminant result, analysis through special item by constitution is shown the high agreement as compared with that through all items.And the discrimination between the special constitution is higher than that between the four constitution.3. The evolution of sound algorithm by Sasang constitution. From a total error rate of view, we could know the more excellent hit ratio could be shown by the new applied statistic method, decision tree structure and artificial nerve net model.We could know that, especially, cart algorithm could be most preferred model, because of error rate and simpleness of analysis.On the sound recognition data, dart algorithm calculate the error rate 3.33%. This could be seen not only the simpleness of analysis but also the contribution to the dimensional reduction.

목차 Contents

• 제 1 장 서론...15
• 제 2 장 국내외 기술개발현홍...18
• 제 3 장 연구개발 수행내용 및 결과...20
• 제 1 절 사상체질 분류와 모집단의 개발...20
• 1. 연구개발의 목표...20
• 2. 조사방법...24
• 3. 연구 결과...27
• 4. 고찰...54
• 제 2 절 사상체질 모집단별 음성분석...61
• 1. 연구방법...61
• 1.1 연구대상...61
• 1.2 기구 및 방법...61
• 1.3 통계처리...63
• 2. 결과...64
• 2.1 설문조사 결과...64
• 2.2 시뮬레이션 결과 ("아"와 "사랑합니다")...65
• 2.3 문장의 발음 시간 비교...74
• 2.4 문자 발음 시간 비교...85
• 2.5 "아"에 대한 Formant Frequency, Pitch, Energy 분석...86
• 2.6 체질판별 분석결과...91
• 2.8 "가나마사아" 비교분석...96
• 3. 고찰...102
• 제 3 절 사상체질별 성문 알고리즘 개발...109
• 1. 자료...109
• 2. 분석방법...109
• 3. CART(Classification and Regression Trees)...114
• 4. CART알고리즘을 이용한 여러 분석결과...120
• 5. 결론...124
• 제 4 장 연구개발목표 달성도 및 대외기여도...125
• 제 5 장 연구개발결과의 활용성과 및 계획...126
• 제 6 장 기타 중요변경사항...128
• 제 7 장 참고문헌...129
• 최종보고서 요약(초록)...137