http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
Comparative Study on Vocal Cepstral Emissions of Clinical Depressed and Normal Speakers
Thaweesak Yingthawornsuk 제어로봇시스템학회 2011 제어로봇시스템학회 국제학술대회 논문집 Vol.2011 No.10
The Mel-scale Cepstral Coefficients of the female speakers’ speech sample clinically categorized into depressed and remitted speech groups are computationally estimated using the 16-triangular filter bank, and consequently taken to represent for input vector to selected classifier. The statistical measures such as the Fisher’s discrimination and the separation power ranking calculation are employed on the extracted feature set for observing on their descriptive statistics, probabilities, and discriminant scores. The higher-order cepstral coefficients reveal the significant difference in term of class separation between depressed and remitted speech samples. The results of classification are supportive of the prior statistical measures evaluated on the cepstral acoustic parameters which imply the MFCCs capable of being potential indicator of depression symptom in female speakers compared to normal females.
Thaweesak Yingthawornsuk,Richard G. Shiavi 제어로봇시스템학회 2008 제어로봇시스템학회 국제학술대회 논문집 Vol.2008 No.10
Two types of speech recording collected from three groups of male subjects clinically diagnosed with depression, remission from depression, and suicidal potential were analyzed and investigated for their acoustic features derived from sub-band energy over 0-2 KHz and GMM-based spectrum of the vocal tract response. Spontaneous and text-reading speech samples characterized by different vocal features revealed significant between-class separation power. Especially, features extracted from the reading speech seemed to provide more separability between classes than those of the spontaneous speech. Additionally, high classification accuracy confirmed that the studied features were capable of distinguishing groups of different diagnostic subjects efficiently. In classifying depressed/suicidal subjects the correct score of classification was at 88.5% for features extracted from reading speech samples, while 85.58% was found from classifying spontaneous speech features. These results were considered to be fairly high in classification performance, which is supportive of the promising ability to distinguish two diagnostic groups whose speech samples changed in their acoustic properties and correlated of serious mental states, known as vocal affects. Our findings suggested some clues in diagnosis of psychiatric disorders for psychiatrist.
Thaweesak Yingthawornsuk,Chusak Thanawattano 제어로봇시스템학회 2010 제어로봇시스템학회 국제학술대회 논문집 Vol.2010 No.10
The empirical results of investigating vocal correlate of depression in female adults are presented in that the certain acoustical property of spoken sound based on spectral entropy is capable of relating the affect change in speech with the symptom severity in diagnostic speakers. Studied sub-band entropies achieved the 93% correct classification in classifying two classes of depressed and remitted speech samples with Support Vector Machine (SVM). By validating the entropy feature models modified on the basis of F-ratio measures, the improvement in classification performances is significantly increased.
Feature Selection Consideration for Multi-Class Cardiac Arrhythmia Classification
Chusak Thanawattano,Thaweesak Yingthawornsuk 제어로봇시스템학회 2010 제어로봇시스템학회 국제학술대회 논문집 Vol.2010 No.10
This paper presents the performance of support vector machine to classify the multi-class arrhythmia dataset by pre-selecting sets of feature that best suit the training data set in two-class fashion. By allowing freedom of feature dimension selection in different grouping in classification procedure, the classification performance is comparable to one that uses constant feature dimension but with less computational complexity.
Terapong Boonla,Thaweesak Yingthawornsuk 제어로봇시스템학회 2011 제어로봇시스템학회 국제학술대회 논문집 Vol.2011 No.10
The acoustical properties of speech have been reported to relate to the mental state of speaker while speaking. This proposed work describes way to address the issue of distinguishing between female depressed patients and female remitted subjects based on the measurable change in the cepstral parameters extracted from their sound record. The cepstral coefficients corresponding to the filter response characteristics, affectively mediated by the emotionally depressive illness or even in particular case of the elevated suicidal risk into the speech production system of depressed speaker, are analyzed via the speech cepstral estimation in conjunction with the GMM fitting approximation. The results of pairwise classification in combinations with SVM, cross-validation, training and testing the cepstral coefficients provide the fairly high accuracy in class separation, when evaluating the testing datasets of coefficients extracted from speech segmentations which are highly corresponding to individual female speakers.