听力损失患者之语音鉴别度客观量化分析机制developanobjective.pdfVIP

聽力損失患者之語音鑑別度客觀量化分析機制 Develop an Objective and Quantifiable Measurement to Analyze Speech Discrimination of Hearing Impairment 陳有圳 、陳順入 Yeou-Jiunn Chen 、Shun-Ju Chen 工業技術研究院電腦與通訊研究所 摘要 聽力受損會嚴重影響個人的人際關係、社會行為與學習成長，配戴助聽輔具 可改善聽力受損狀況，而客觀量化聽力損失患者的語音鑑別度研究，則可協助聽 輔具的研發、調整與聽能復健。在本論文中提出自動化的聽力損失患者之語音鑑 別度客觀量化分析機制，可克服臨床診療中純音聽力檢查中，只能評估受測者的 聽力偵測閥值無法表達語音鑑別度的問題，在系統中透過聽力損失模擬機制與連 續語音樣型比對技術的整合，能避免主觀式詞彙聽力測試中受測者年齡限制、語 言環境等客觀環境限制，並降低冗長聽力測試造成的測試結果不確定性，更能比 只考慮頻率能量分佈的發音索引技術，進一步模組人類對於語音訊號的理解認知 特性。 ABSTRACT With the loss of hearing, man is restricted from his or her normal social activity. The hearing loss will lead to decrease the speech discrimination in syllable or phoneme recognition. In the clinical protocol, subjective hearing test is adopted to assess the speech discrimination. But, this approach is hard to perform a lot various kinds of experiments, and even worse, their responses are not reliable from time to time. Moreover, the Articulation Index is applied to objectively estimate the speech discrimination. However, the energy in each frequency channel was considered in the Articulation Index and cannot evaluate the speech perception of man. In this paper, an objective and quantifiable measurement was proposed to analyze the speech discrimination of hearing impairment. Applying hearing impairment simulation, continuous word pattern recognition, and statistical analysis technique, the speech discrimination of a specified hearing impairment can be analyzed. Experimental results showed that the simulated speech discrimination and that of subjective hearing test are consistent in statistical analysis. Moreover, the proposed system