Gait Analysis Using Principal Component Analysis and Long Short Term Memory Models

Abstract

Human analysis and diagnosis have become attractive technology in many fields. Gait defines the style of movement and gait analysis is a study of human activity to inspect the style of movement and related factors used in the field of biometrics, observation, diagnosis of gait disease, treatment, rehabilitation, etc. This work aims in providing the benefit of analysis of gait with different sensors, ML models, and also LSTM recurrent neural network, using the latest trends. Placing the sensors at the proper location and measuring the values using 3D axes for these sensors provides very appropriate results. With proper fine-tuning of ML models and the LSTM recurrent neural network, it has been observed that every model has an accuracy of greater than 90%, concluding that LSTM performance is observed to be slightly higher than machine learning models. The models helped in diagnosing the disease in the foot (if there is injury in the foot) with high efficiency and accuracy. The key features are proven to be available and extracted to fit the LSTM RNN model and have a positive outcome.