Transferring speech-generic and depression-specific knowledge for Alzheimer's disease detection
This work addresses data sparsity in Alzheimer's disease detection for healthcare applications, but it is incremental as it builds on existing knowledge transfer methods.
The paper tackles the issue of sparse training data for Alzheimer's disease detection from speech by transferring knowledge from speech-generic foundation models and a depression detection task, achieving a state-of-the-art F1 score of 0.928 on the ADReSSo dataset.
The detection of Alzheimer's disease (AD) from spontaneous speech has attracted increasing attention while the sparsity of training data remains an important issue. This paper handles the issue by knowledge transfer, specifically from both speech-generic and depression-specific knowledge. The paper first studies sequential knowledge transfer from generic foundation models pretrained on large amounts of speech and text data. A block-wise analysis is performed for AD diagnosis based on the representations extracted from different intermediate blocks of different foundation models. Apart from the knowledge from speech-generic representations, this paper also proposes to simultaneously transfer the knowledge from a speech depression detection task based on the high comorbidity rates of depression and AD. A parallel knowledge transfer framework is studied that jointly learns the information shared between these two tasks. Experimental results show that the proposed method improves AD and depression detection, and produces a state-of-the-art F1 score of 0.928 for AD diagnosis on the commonly used ADReSSo dataset.