Abstract: Stress is an important prosodic feature for stress-timed language such as English. This paper presents a new approach using auditory model-based Pitch Synchronous Peak Amplitude(PSPA) feature, which incorporates frequency and intensity information in English lexical stress detection. PSPA feature, along with traditional features and their combinations to English lexical stress detection are evaluated with ISLE database. Experimental results show that the combination of new feature and traditional features demonstrates a 1.5% error rate reduction.

Key words: stress detection, auditory model, Pitch Synchronous Peak Amplitude(PSPA)

摘要： 对于英语等“重音节拍语言”，重音是一个非常重要的韵律学特征。从听觉模型的角度出发，利用基音同步幅度峰值特征能同时表征瞬时频率和强度信息的特点进行重音检测。使用基音同步幅度峰值特征以及与传统特征的组合对英语连续语音的试验结果表明，新特征能使系统误识率降低1.5%。

关键词: 重音检测, 听觉模型, 基音同步幅度峰值

CLC Number: 

• TN912.34