Estimation Model for Emotions Based on Pulse

The progressive aging of society has increased expectations for the spread of nursing care robots to support long-term care and welfare services. This research had the goal of developing a communication system as one of the elemental technologies of nursing care robots, along with a method that allows care robots to consider a user’s emotions. The estimation of emotions based on a user’s electroencephalogram and heartbeat has attracted attention. However, users may experience stress when wearing the sensors needed for such measurements. To prevent this system from causing stress in users, we had the goal of developing an estimation model for emotions based on the pulse, which is relatively easy to measure. Various autonomic nervous activity indices (pNN50, RMSSD, LF, HF, LF/HF) were adopted for the estimation model, and transfer functions were established. These indices were considered in time domain and frequency domain analyses of the heart rate variability. The pulse was measured while the user was watching a video and converted into an accelerated plethysmogram using second order differentiation. Then, the autonomic nervous activity indices were calculated. The transfer function from the input to output was identified using these autonomic nervous activity indices as inputs and the responses to a questionnaire that was administered after watching the video as outputs.