Speech entrainment and how acoustic complexity drives attention
Pierre Labendzki’s research aims at bridging the gap between two fields of research : speech entrainment and how acoustic complexity drives attention. Speech entrainment is the ability of the auditory cortex to ”track” speech amplitude modulations. This ability is thought to enhance speech perception but fewer is known for infants, especially in naturalistic environment. We research how and why infants brains entrain to the caregiver speech envelope. Attention scaffolding by the caregivers is key to the infant language acquisition. Infant directed speech (IDS) differs from adult directed speech (ADS) in many acoustic and lexical ways. Previous research tends to show that the fundamental frequency on IDS is less predictable than in ADS, presumably for attention grabbing. On the other hand, IDS envelope is slower and more predictable and in ADS, we will try to show why. I use different metrics to assess auditory complexity, such as a lossless compression algorithm [1] for general complexity and long short term memory neural network for F0 and envelope predictability.
Figure 1: Complexity over time, where 1 means max complexity, the method used is lossless compression as proxy for Kolmogorov complexity. The score is Veridisco by Daft Punk (2001)
Background
I obtained a master degree in engineering at Paris East University. My engineering dissertation (CNRS/ INCC BabyLab, Paris) focused on infant amplitude modulation perception and speech in noise listening abilities. I also completed a master in music informatics at the Gustave Eiffel University, where my dissertation addressed the complexity of musical pieces.
References
- Labendzki, P. MetricForComplexity, GitHub repository of the project, https://github.com/pierrelabendzki/MetricForComplexity , 2021.