Learning Domain Invariant Representations for Child-Adult Classification from Speech

TL;DR

This paper develops a domain adversarial learning approach to improve child-adult speaker classification in speech analysis for autism diagnostics, addressing variability due to age and data source, and demonstrates significant performance gains.

Contribution

It introduces domain adversarial training methods to create speaker embeddings invariant to age and data source, enhancing classification robustness without labeled target data.

Findings

Up to 13.45% relative improvement over conventional methods.

Effective handling of variability due to age and data source.

Demonstrated on large autism diagnostic dataset.

Abstract

Diagnostic procedures for ASD (autism spectrum disorder) involve semi-naturalistic interactions between the child and a clinician. Computational methods to analyze these sessions require an end-to-end speech and language processing pipeline that go from raw audio to clinically-meaningful behavioral features. An important component of this pipeline is the ability to automatically detect who is speaking when i.e., perform child-adult speaker classification. This binary classification task is often confounded due to variability associated with the participants' speech and background conditions. Further, scarcity of training data often restricts direct application of conventional deep learning methods. In this work, we address two major sources of variability - age of the child and data source collection location - using domain adversarial learning which does not require labeled target domain data. We use two methods, generative adversarial training with inverted label loss and gradient reversal layer to learn speaker embeddings invariant to the above sources of variability, and analyze different conditions under which the proposed techniques improve over conventional learning methods. Using a large corpus of ADOS-2 (autism diagnostic observation schedule, 2nd edition) sessions, we demonstrate upto 13.45% and 6.44% relative improvements over conventional learning methods.