Assessing the effects of fundamental-frequency dynamics on the intelligibility of competing voices

Young listeners with a healthy auditory system are capable of understanding speech in the presence of one or several interfering voices, even in the most challenging listening scenarios. This ability is crucial for daily social life, but it is compromised in older listeners affected by sensorineural hearing loss, who often experience difficulties understanding speech in complex auditory environments, even when hearing-aid solutions are provided. Investigating auditory scenarios with multiple speakers is thus essential for revealing phenomena that can inspire the development of hearing-loss compensation strategies.
Among many auditory cues, the fundamental frequency (F₀) and its differences between competing voices provide useful information that aids target speech intelligibility, which can be successfully utilized by normal-hearing (NH) listeners, whereas older hearing-impaired (HI) listeners have limited access to it. Evidence for the efficacy of F₀ information has been obtained by means of laboratory simulations of competing-talker scenarios, with highly constrained speech stimuli that are not truly representative of the characteristics and the variety of realistic speech and therefore do not guarantee the reproducibility of the results in the diversity of auditory situations encountered in daily life.
This thesis aims to expand the available knowledge on the role of F₀-related cues in competing-talker scenarios by using naturalistic everyday speech stimuli with a wide variability of F₀ characteristics that is typical of realistic voices. The effects of differences in average F₀ and F₀ dynamic range between competing voices on speech perception were investigated in NH and HI listeners. For NH listeners, the measured effects of these cues on speech intelligibility were small or negligible. The average F₀ difference between competing voices was found to only provide a speech-intelligibility benefit when energetic cues were limited or absent, which occurs especially when the competing voices have unrealistically similar syntactical structure and F₀ trajectories. The effect on speech intelligibility induced by the difference in F₀-dynamic-range between competing voices was found to be negligible. However, it was shown that the presence of a relatively large F₀ dynamic range in at least one of the two competing sentences improved speech intelligibility, regardless of the difference in F₀ dynamic range between sentences. For HI listeners, the inability to utilize these F₀ cues was confirmed: compared to NH listeners, the benefit induced by an average F₀ separation between competing voices was smaller and no significant effect of F₀-dynamic-range of the sentences nor of their difference was observed. Finally, an analysis of the F₀ properties of speech recordings from naturalistic dialogues was presented, providing a reference for the F₀ properties of realistic speech and describing the changes in F₀ properties that talkers produce in the presence of communication barriers such as background noise and hearing impairment.
This thesis contributes to the body of literature on the role of F₀-related cues in communication scenarios, by proposing new methodologies that focus on the realism of the speech materials and on the numerical control of the experimental method. Overall, the results of this thesis suggest that in realistic competing-talker scenarios, the F₀-related cues contribute to a holistic picture of the auditory scene that involves many auditory cues. In such scenarios, especially the F₀ dynamic range of the individual competing sentences can affect speech intelligibility.