This paper is concerned with voice source variation associated with different emotional portrayals of an utterance: bored, sad, happy, surprised, angry and neutral. The source analyses involved pulse-by-pulse inverse filtering to yield the differentiated glottal flow, and subsequent parameterisation of the source signal using the LF model. The glottal source parameters included in the analysis were F0, EE, RK, RG, RA, FA, OQ and RD. For the data set analysed, each emotion seems to have its own distinct pattern of source parameter settings. Analysis of the dynamics of the source variation illustrated here on the RD parameter suggests that to better understand source variation we need to study it in terms of the prosodic components of the utterance.

Index Terms. Voice source, dynamics, emotion