암밴드형 근전도 센서를 이용한 앙상블 인공 신경망 학습 기반의 실시간 지화 인식 시스템

다국어 초록 (Multilingual Abstract)

Deaf people using sign language and finger language experience social inequalities and financial losses due to communication restrictions. This study developed a finger language recognition algorithm based on ensemble artificial neural network(E-ANN) using an armband system with multi EMG sensors. The developed algorithm was composed of signal acquisition, filtering, segmentation, feature extraction and E-ANN based classifier. The algorithm was evaluated from 17 subjects. 80% of the feature vectors obtained from Korean finger language(14 consonants, 17 vowels, 7 number) were used for classifier training and 20% for evaluation. The E-ANN consisting of classifiers from 1 to 10 was trained using 50∼1500 training data. The accuracy of all classifiers was obtained through 5-fold cross validation and the accuracy of the optimal E-ANN and general ANN were compared. With the number of both training data and classifiers the mean accuracy of the classifier increased and SD decreased. Beyond the 300 training data or 8 classifiers, the accuracy of the algorithm was not improved any more but significantly higher than that of general ANN.

목차 (Table of Contents)

• ABSTRACT
• 서론
• 연구 방법
• 결과
• 고찰 및 결론
• 참고문헌