Perception of the Fundamental Frequency of Continuous Speech

Abstract

Three tape recorded samples of sentence material having average fundamental frequencies of 119, 139, and 179 Hz were selected from a larger stimulus set [J. F. Brandt, “Effects of Stimulus Bandwidth on Listener Judgments of Vocal Loudness and Effort,” J. Acousl. Soc. Am. 52, 705–707 (1972)]. The speech samples were presented to five normal-hearing listeners, unfiltered, and high-pass and low-pass filtered at 1500 Hz. Pitch matches to the speech samples were made using a filtered ac pulse train of varying repetition rate and energy centered around 2200 Hz. The fundamental frequency of the speech samples was perceived by the listeners in basically the same way that periodicity pitch (nonspeech signals) information is obtained. That is, the fundamental pitch information contained in continuous speech can be easily obtained from temporal (envelope) information in the physical absence of low-frequency energy. Listeners could also recognize differences between normal-speaking fundamental frequencies and fundamental frequencies produced at frequencies higher and lower than normal.