Exploring conversational entrainment in noise using Cross-Recurrence Quantification Analysis (CRQA)

Conversational partners often adjust aspects of their speech to one another, a phenomenon known as entrainment. Because such alignment is thought to support mutual understanding and smooth turn‑taking, examining how it behaves in adverse listening conditions is important for understanding communication challenges faced by many listeners.

In this work, we investigate how background noise affects the dynamical structure of f₀ in conversation using a nonlinear, trajectory‑based Cross‑Recurrence Quantification Analysis (CRQA) framework. Using data from 24 older Danish adults completing the Diapix task in quiet, 60 dBA, and 70 dBA SPL multi‑talker babble noise, we extracted continuous f₀ trajectories and used CRQA to quantify shared patterns between interlocutors.

This analysis highlights how noise reshapes individual f₀ behavior and how those changes propagate into CRQA metrics. Our preliminary results show that noise‑related reductions in recurrence primarily reflect altered individual f₀ structure rather than weakened coupling between speakers. The poster discusses key methodological considerations in CRQA for conversational f₀: embedding optimization, delay selection, normalization choices, and handling of silences, and it demonstrates how these decisions influence interpretations of recurrence under noisy conditions.

These findings provide an initial methodological foundation for applying CRQA to conversational prosody in difficult listening environments and motivate future work on speech entrainment in populations with hearing impairment.