ANN-Based Human Activity Recognition (HAR)

Project Brief

This project involves creating a custom-made backpropagation neural network from scratch using Python for human activity recognition based on motion sensors and gyroscope time-series data. The implementation includes comprehensive exploratory data analysis, data preprocessing, and extensive parameter tuning including optimization of neuron counts, layer configurations, activation functions, and backpropagation learning mechanisms. The project achieved exceptional results with 98% accuracy in recognizing human activities from sensor data.

Skills & Technologies