Predictability of Maternal Sensory Signals: A Machine Learning Perspective

Mother-infant interactions have a significant impact on early childhood development, including emotional, social, and cognitive development. These interactions foster secure attachments, improve language skills, and develop social competencies. The typically used method for analyzing maternal predictability relies on video observations of mother-infant interaction and manual coding by trained psychologists; a method that is accurate but labor-intensive and costly. Recent advances in the Machine Learning (ML) and Deep Learning (DL) technologies provide possibilities for more efficient and dynamic methods of assessing mother-infant interaction. This study introduces an advanced approach, employing a transformer-based model to predict tactile stimuli, Voice Activity Detection (VAD) for auditory stimuli, and Computer Vision techniques to analyze visual stimuli, focusing on gaze direction detection. By integrating these multimodal analysis methods, we compute a comprehensive measure of maternal behavioral predictability, quantified through the calculation of entropy rate across the observed sensory signals. Our model achieves 74.8% accuracy in identifying and categorizing tactile stimuli, with auditory and visual stimuli analyses still under evaluation. This initial success demonstrates the potential of our multimodal approach, while comprehensive performance assessment across all sensory domains is ongoing. This study highlights the transformative potential of technology-driven methods in developmental psychology research.