Perceptual experiment and acoustic analysis of Chinese attitudes: A preliminary study

The present work studied 13 types of Chinese attitudes and compared the confusion patterns between subjective and objective identification. The listening experiment showed that the overall recognition rate for these attitudinal utterances was 46% by native subjects, while in discriminant analysis, the recognition rate was 25.4% based on five prosodic parameters (minimum f0, maximum f0, mean f0, f0 standard deviation, and speaking rate). Cluster analysis of the subjective and objective confusion patterns showed some similarities between them. For example, “friendly” and “polite”, “hostile” and “rude” were confused in both subjective and objective identification. However, some attitudes were prosodically similar, e.g., “neutral” and “sincere”, but the subjects were able to distinguish them. On the contrary, neutral utterances were prosodically different from submissive utterances, while subjects still confused large amount of submissive utterances to “neutral”.