InfantNet: A Deep Neural Network for Analyzing Infant Vocalizations

Abstract

Acoustic analyses of infant vocalizations are valuable for research on speech development as well as applications in sound classification. Previous studies have focused on measures of acoustic features based on theories of speech processing, such spectral and cepstrum-based analyses. More recently, end-toend models of deep learning have been developed to take raw speech signals (acoustic waveforms) as inputs and convolutional neural network layers to learn representations of speech sounds based on classification tasks. We applied a recent endto-end model of sound classification to analyze a large-scale database of labeled infant and adult vocalizations recorded in natural settings outside the lab with no control over recording conditions. The model learned basic classifications like infant versus adult vocalizations, infant speech-related versus nonspeech vocalizations, and canonical versus non-canonical babbling. The model was trained on recordings of infants ranging from 3 to 18 months of age, and classification accuracy changed with age as speech became more distinct and babbling became more speech-like. Further work is needed to validate and explore the model and dataset, but our results show how deep learning can be used to measure and investigate speech acquisition and development, with potential applications in speech pathology and infant monitoring.

Emerging Verbal Functions in Early Infancy: Lessons from Observational and Computational Approaches on Typical Development and Neurodevelopmental Disorders

“…infant vs. adult, canonical vs. non-canonical). However, none of these studies focused on infants with DDs in the first few months of life ( Ebrahimpour et al, 2020 ; Warlaumont et al, 2010 ).…”

“…spectrograms, waveform, parametric representation), and classification schemes (i.e. infant-directed speech vs. adult-directed speech, infant vs. adult, vocalisation vs. non-vocalisation, canonical vs. non-canonical; vocant vs. squeal vs. growl; Ebrahimpour et al, 2020 ; Li et al, 2021 ; Warlaumont et al, 2010 ). Opposed to that, in the babbling phase, a number of studies analyse verbal capacities utilizing computational approaches (e.g.…”

“…These perceptually less salient vocalisations, as compared to canonical babbling, have preferably been studied through simple LLDs such as F 0 and duration. Only a few studies have used more advanced computational approaches to prove the applicability and value of such approaches in the field of pre-babbling vocalisations ( Ebrahimpour et al, 2020 ; Li et al, 2021 ; Warlaumont et al, 2010 ). Besides missing analytical approaches, there is also a lack of standardisation of coding-schemes and datasets, which impedes the comparability of performance between applied computational models in the field of speech-language analysis in the first 5 months of life.…”

“…In addition, machine learning models have been applied to vocalisations including neural networks in different varieties, testing classification tasks (e.g. adult vs. infant speech, canonical vs. non-canonical utterances; Ebrahimpour et al, 2020 ; Warlaumont et al, 2010 ). In our group, we have utilised a machine learning approach (i.e.…”