Utterance Condition Determination via Personalized Backchannel Frequency

The system performs preliminary action by estimating each speaker's average backchannel frequency before the actual emotional condition determination. This preliminary estimation creates a personalized baseline for each speaker, allowing the system to account for individual variations in backchannel feedback patterns. The average backchannel frequency is calculated over a predetermined period and stored as reference data, which is then used to normalize and compare actual backchannel frequencies during emotional state detection, thereby improving determination accuracy despite individual differences.

The system applies local quality by transitioning from a uniform fixed threshold applied to all speakers to personalized dynamic thresholds based on each speaker's estimated average backchannel frequency. Each speaker receives a customized determination threshold derived from their individual communication patterns. This localized approach maintains operational simplicity through automated personalization while significantly improving measurement precision by adapting to individual variations in backchannel feedback behavior.

The system implements self-service by automatically estimating each speaker's average backchannel frequency and generating personalized determination thresholds without requiring manual configuration or intervention. The system autonomously adapts to each speaker's individual patterns by processing their communication history and computing their unique baseline metrics. This self-configuration capability improves measurement precision through personalization while minimizing the increase in operational complexity, as the system performs these calculations automatically during normal operation.