Prosodic Feature Criterion for Hebrew Using Different Feature Sets

摘要

Prosody is essential for everyday human communication and provides important information about intention and meaning. It is used for subtle expressions such as sarcasm as well as for denoting more common expressions like questions or declarations and even can indicate the physiological or emotional condition of a speaker. In our previous work we presented a Prosodic Feature Criterion (PFC) for evaluating the prosodic nature of a feature that was extracted from speech signal. The PFC score provides us with a way to rank the features and determine whether an acoustic or spectral feature carries prosodic information. In this paper we continue to explore this mechanism, using the OpenSMILE toolkit, which is a standard set of features widely used for acoustic analysis and prosody research. Our experiments are carried out using a dataset of Hebrew utterances specifically designed for prosody research. We apply the PFC over each feature separately, thus ranking the different features. We then compare this ranking with classification based ranking of the same features. In addition we show visualization of the PFC idea using dimension reduction of multiple features representation. Both these tests, validate the use of the PFC score, for evaluating the prosodic nature of a feature in regards to specific prosody classes.