RAHUL ANANTMathematics & Computing
Back to Projects Library
Healthcare AI & Sensor FusionCompleted2026

Project SAMARTH: Kinematic AI & Rehabilitation System

Role: Machine Learning Intern

AI-powered physiotherapy and rehabilitation system for lower-limb movement assessment developed under faculty mentorship at IIT (BHU).

Verified Project Links & Repositories
1 Active Resource
Project SAMARTH: Kinematic AI & Rehabilitation System

Technology Stack & Libraries

PythonPyTorchOpenCVMediaPipeNumPyPandasScikit-learn
The Challenge / Problem

Context & Objectives

Clinical gait assessments traditionally require costly optical motion laboratories or subjective visual scales.

Architectural Solution

Engineered Approach

Combined wearable 9-DOF IMU data with MediaPipe markerless computer vision to analyze exercise quality, range-of-motion, and symmetry.

System Architecture & Pipeline

Wearable IMUs + MediaPipe Pose -> Kinematic Feature Extraction (44 dims) -> Zero-Overlap Windows -> BiLSTM-Attention & STGCN-LSTM Autoencoders -> Movement Scoring.

Built an AI-powered physiotherapy and rehabilitation system for lower-limb movement assessment under faculty mentorship at IIT (BHU). Developed and validated ML pipelines combining wearable IMU sensor data and human pose estimation (MediaPipe) across the MHEALTH and MoVi datasets.

Engineered Capabilities & Innovations

44-dimensional kinematic feature extraction across MHEALTH and MoVi datasets
BiLSTM-with-attention, 1D CNN, and STGCN-LSTM autoencoder models
Zero-overlap windowing and Leave-One-Subject-Out (LOSO) cross-validation
Real-time range-of-motion, repetition count, and symmetry analysis

Empirical Results & Benchmarks

Featured on the live team at Physio Samarth (physiosamarth.com/our-team) supervised by Dr. Pankhuri Jain at IIT (BHU).

Have Questions About This System?

I am always glad to discuss technical architecture, benchmarks, or potential collaboration.