Speech Recognition for Aphasia Treatment Based on OpenEars

• Main Authors: You-sheng Shiu, 徐有盛
• Other Authors: San-Yih Hwang
• Format: Others
• Language: en_US
• Published: 2013
• Online Access: http://ndltd.ncl.edu.tw/handle/64266535606507881740

碩士 === 國立中山大學 === 資訊管理學系研究所 === 101 === In Taiwan, there are thousands of aphasic patients who need professional rehabilitation service. The shortage of therapists causes two problems: 1) not all patients can make clinic appointment at the desired time and 2) the average length of rehabilitation session is reduced. These patients need a way to satisfy the need and improve the quality of aphasic treatment. Dephasia is a computer system for aphasic rehabilitation based on Web and iPad. In this thesis, we proposed a way to provide real-time evaluation for Chinese speech input in aphasic treatment on the client-side for two types of questions, namely repeating sentence and naming practice. This extension is based on OpenEars, a free shared-source SDK for iPhone Voice Recognition and text to speech. We propose some designs and modification to adapt the system to aphasic treatment. We evaluate our proposed approach using 156 samples from 10 different patients. The evaluation is based on accuracy and efficiency running on iPad2, iPad4 and MBP. The result shows that for ordinary recordings, the Pearson correlation between scores given by proposed solution and the therapists could reach up to 0.59, and the recognition time is within 5 seconds. The result shows that it is possible to apply speech recognition to speech therapy to provide real-time feedback on client-side.