Mutual information between heart rate variability and respiration for emotion characterization

Area of Science:

• Physiology and Biomedical Engineering
• Affective Computing and Human-Computer Interaction

Background:

• Emotion recognition is crucial for diagnosing psycho-neural illnesses.
• Heart rate variability (HRV) and respiratory signals contain complex information related to emotional states.
• Advanced signal processing techniques are needed to extract meaningful patterns from physiological data.

Purpose of the Study:

• To investigate the utility of auto-mutual information function (AMIF) and cross-mutual information function (CMIF) for human emotion recognition.
• To analyze complex interdependencies and couplings within HRV and between HRV and respiratory signals.
• To evaluate the ability of these information-theoretic measures to discriminate between different emotional states.

Main Methods:

• Applied AMIF to heart rate variability (HRV) signals and CMIF to quantify coupling between HRV and respiratory signals.
• Adapted algorithms for short-term RR time series, including band-pass filtering (LF, HF, and respiration-based).
• Calculated AMIF and CMIF at various time scales to derive complexity measures for emotion discrimination.

Main Results:

• AMIF on filtered RR time series significantly discriminated between neutral (relax), joy, fear, sadness, and anger states (p < 0.05, accuracy > 70%, AUC > 0.70).
• AMIF and CMIF parameters effectively characterized the low signal complexity associated with fear compared to other emotions.
• Demonstrated the potential of information-content complexity derived from physiological signals for emotion classification.

Conclusions:

• Human emotions can be effectively characterized by analyzing the information-content complexity of HRV and respiratory signals.
• AMIF and CMIF are promising tools for objective emotion recognition in physiological data.
• This approach holds potential for advancing diagnostic tools for psycho-neural conditions.