Piezoelectric continua modeling with non-linear electromechanical coupling for ASD signal classification

Autism Spectrum Disorder (ASD) is a neurodevelopmental disorder with atypical motor skills and repetitive movements. Current diagnostic methods, such as electroencephalogram (EEG) and video-based systems face challenges, including the limited interpretability of the signals, computational complexity, and low real-time biomechanical sensitivity.To overcome these limitations, this article presents a wearable biomechanical sensing system using non-linear electromechanical modeling of piezoelectric continua to achieve efficient signal sensing and ASD prediction.The model describes the non-linear dynamic electromechanical behavior of a piezoelectric-based system allowing for high-sensitivity sensing of both macro- and micro-movements.Finite element analyses show that a voltage variation from 4.1 mV to 80.7 mV is generated for an applied force of 0.5 N and 10 N, which confirms the electromechanical sensitivity of the sensor.Multiscale signal analysis shows ASD-related micromotions mostly occur at frequencies ranging from 3–8 Hz, with a rate of 72% (versus 28% in control group). A hybrid PVDF-PZT design for the sensor system has a sensitivity of 14.1 mV/N and signal to noise ratio of 31.6 dB.A machine learning-based classification with ensemble model attains 96.2% accuracy and real-time detection using the model is 95.7% accurate.The developed approach offers a scalable, non-contact and highly accurate scheme for ASD diagnosis through physics-based and data-driven approaches.