Voiceless Intention Log: 頸部生体情報を用いた 発話欲求推定に基づく会話参加支援システム

In group conversations, situations arise where individuals may wish to speak but are unable to do so due to the flow of the conversation or constraints on the timing of their turn. While previous studies have proposed methods for estimating the desire to speak and turn-taking using posture changes, eye movements, and respiration, few studies have focused on physiological signals measured from the neck, which contains key structures involved in speech production. In this study, we propose a conversation participation support system that estimates physiological responses related to the desire to speak by analyzing changes in blood flow, pulse waves, and acceleration data obtained from XB-01, a compact brain activity measurement device worn on the neck. It has been reported that physiological responses preparatory to speech, such as respiratory adjustments and laryngeal muscle activity, occur prior to the onset of speech. Therefore, by comprehensively analyzing changes in blood flow and body movements in the neck, we attempt to detect states in which a person has the desire to speak but has not yet begun speaking. Furthermore, by visually signaling the estimated results to those around the user, we aim to support the coordination of speaking opportunities and facilitate smooth communication.