Design and Validation of a System to Synchronize Speech Recognition and Eye-Tracking Measurements

Dementia and other forms of cognitive decline can cause aging adults to have to leave their homes and enter care facilities. These declines can affect eye movements and speech patterns. This paper proposes a system that combines and synchronizes speech recognition and eye-tracking analysis for early-stage Alzheimer's diagnosis through a computer-based analysis system. The goal is to enable the combination of these two measures within a system such as a machine learning model that can simultaneously analyze multiple types of data. This can improve the prediction performance compared to previous studies. However, before building and training the model, it is crucial to synchronize the speech and eye-tracking data streams accurately. This paper focuses on the design and validation of the proposed synchronization task to align eye tracking results from a gaze tracking system with spoken words identified by a speech recognition system. The synchronization task involves measuring the relative and absolute delay between two vision tracking systems: Tobii Eye-tracker, Webcam Eye-tracker, and recognized speech stream. A pilot sample of 9 subjects was used to validate the synchronization task. The study recorded 90 samples, and the results show that the proposed task is feasible for accurately synchronizing the data streams.