Speech and Breath-Sound-Based Person Identification with Sparse Training Data

• Main Author: TRAN VAN THUAN
• Other Authors: Tsai,Wei-Ho
• Format: Others
• Language: en_US
• Published: 2019
• Online Access: http://ndltd.ncl.edu.tw/handle/tpzt27

碩士 === 國立臺北科技大學 === 電資國際專班 === 107 === This study aims to develop a person identification (PID) system based on combined use of bronchial breath sounds and speech signals acquired by stethoscope. Two major methods, including support vector machines, and artificial neural networks are evaluated in the task of breath-sound-based PID. With the consideration of convenience, the amount of sound data collected from each person should be as small as possible, and hence the performance of PID may be limited, when the sound data for training the system is insufficient. To boost the performance of PID with sparse training data, this work studies data augmentation (DA) techniques that avoid the system training process from the overfitting problem. In addition, Feature engineering techniques are utilized to find the informative subset of breath sound features which is beneficial for PID. Our experimental data is provided by 16 volunteers, including equal number of male and female participants. In test phase, Both Support vector machine in combined with feature selection and Artificial Neural Networks approaches yielded the comparable accuracies of 98%.