Science Academy Press

Search:

IJMLIP - Machine Learning & Image Processing

Extended Kalman Filter to Detect Human Body Movements through Inertial and Magnetic Sensors - Medical Applications

Dr. T. Senthil Kumar, Leo John Baptist Andrews

International Journal of Machine Learning and Image Processing, Vol 1, Issue 1, April 2023

Recent Articles from IJMLIP

Open Access Article Machine Learning Image Processing

Extended Kalman Filter to Detect Human Body Movements through Inertial and Magnetic Sensors

Dr. T. Senthil Kumar, VIT University, Vellore | Leo John Baptist Andrews, Botho University, Botswana

Inertial orientation tracking is investigated for real-time human movement capture. This research presents a quaternion-based Extended Kalman Filter (EKF) approach for three-dimensional orientation determination using IMU sensors. The proposed E2QKF system achieves remarkable accuracy with Euler angle errors within 5 degrees, demonstrating effective performance for long-term human body motion measurement with maximum errors less than 3.0 degrees.

Keywords: Extended Kalman filter, inertia, magnetic sensing, orientation, medical applications

Open Access Article Deep Learning Face Detection

Mathematical Modelling of Multi-Task Cascaded Convolutional Networks: Face Detection Application

Dr S B Mohan, Associate Professor, ECE, S.A. Engineering College

A novel deep cascaded multi-task CNN structure for face detection is proposed, utilizing the intrinsic connection between detection and alignment tasks. The approach employs three carefully designed deep convolutional networks in a cascaded architecture, achieving superior accuracy on challenging FDDB and WIDER FACE benchmarks while maintaining real-time performance. The method demonstrates significant improvements over state-of-the-art approaches.

Keywords: face detection, mathematical modeling, convolutional neural networks, deep learning

Open Access Article Machine Learning Prediction

Optimization Techniques and Hybrid Deep Learning Approaches for UV Index Predictions

Dr. A. Rajasekar, Dhaanish Ahmed College | Dr. B.R. Tapas Bapu, S.A. Engineering College | Dr. N. Ashokkumar, Mohan Babu University

This research develops and analyzes hybrid machine learning algorithms combining CNN and LSTM (CLSTM) for daily UV index forecasting in India. The proposed modified CEEMDAN-CLSTM model demonstrates excellent prediction capacity with low error rates and exceptional performance levels. GoogleNet achieved 99% accuracy with the lowest loss rate (0.1557) and shortest runtime (4.14 minutes), proving beneficial for developing practical computer-aided diagnostic frameworks.