Zero-Shot Cognitive Impairment Detection from Speech Using AudioLLM

TL;DR

This paper introduces a zero-shot speech-based method for detecting cognitive impairment using AudioLLM, which performs comparably to supervised models and generalizes well across languages and datasets.

Contribution

It is the first to apply a zero-shot approach with AudioLLM for cognitive impairment detection from speech, eliminating the need for manual annotation.

Findings

Achieves performance comparable to supervised methods

Demonstrates strong cross-lingual and cross-dataset generalization

Works effectively on multilingual and multi-task datasets

Abstract

Cognitive impairment (CI) is of growing public health concern, and early detection is vital for effective intervention. Speech has gained attention as a non-invasive and easily collectible biomarker for assessing cognitive decline. Traditional CI detection methods typically rely on supervised models trained on acoustic and linguistic features extracted from speech, which often require manual annotation and may not generalise well across datasets and languages. In this work, we propose the first zero-shot speech-based CI detection method using the Qwen2- Audio AudioLLM, a model capable of processing both audio and text inputs. By designing prompt-based instructions, we guide the model in classifying speech samples as indicative of normal cognition or cognitive impairment. We evaluate our approach on two datasets: one in English and another multilingual, spanning different cognitive assessment tasks. Our results show that the zero-shot AudioLLM approach achieves performance comparable to supervised methods and exhibits promising generalizability and consistency across languages, tasks, and datasets.