Comparative Analysis of CNN-based Deep Learning Approaches on Complex Activity Recognition
Sakorn Mekruksavanich, Ponnipa Jantawong, Anuchit Jitpattanakul · 2022 Joint International Conference on Digital Arts, Media and Technology with ECTI Northern Section Conference on Electrical, Electronics, Computer and Telecommunications Engineering (ECTI DAMT & NCON) · 2022
For those with Parkinson's disease, freezing of gait (FOG) could be a distressing condition. In individuals with diabetes, FOG could have a significant influence on their quality of life and cause falls, which can be dangerous. A questionnaire has traditionally been used to measure the severity of FOG symptoms; nevertheless, this technique is subjective and can not accurately reflect the intensity of this illness. The adoption of sensor-based devices may give reliable and objective information to follow the progression of symptoms, allowing for better Parkinson's disease management and therapy. A compact deep convolutional neural network comprising squeeze and excitation components, which we called SE-DeepConV model, was designed to identify FOG utilizing wearable data from the sensors. The SE-DeepConV model was constructed and evaluated using a publicly accessible standard FOG dataset, particularly the Daphnet dataset. The SE-DeepConV model outperforms other benchmark deep learning models in terms of accuracy, with the most outstanding accuracy of 95.65%, according to research observations.