Multiple equipment and activity classification using audio signals created by the equipment through deep learning in construction sites

Status: Ongoing

This work demonstrates to recognition of the multiple heavy equipment and their corresponding activities in the construction sites using sound they created based on the mel-spectrogram and Convolutional Neural Network (CNN). Specifically, herein I used a pre-trained ResNet-18 model including several audio signal augmentation techniques such as adding Gaussian noise, pitch shifting, phase shifting, normalizing, etc. The results of the study showed 99% accuracy for equipment classification and 98% accuracy for activities classification.

Simulation-Based Analysis of Permanent Magnet Synchronous Machines

10th Semester, Department of EEE, Southeast University, Dhaka, Bangladesh

This research work demonstrates the simulation of field-oriented control of PMSM. The main reason for this is that, in field-oriented control, both torque and speed can be controlled independently by two currents responsible for torque and flux controlling separately. The whole system is simulated based on the mathematical model of PMSM and field-oriented control method with designed PI controllers.

Line Follower Restaurant Robot by Using Arduino

9th Semester, Department of EEE, Southeast University, Dhaka, Bangladesh

During the undergraduate program, a line follower robotic research, and project was carried out. It was also displayed at an electrical and electronic project fair conducted by the department of EEE at Southeast University in Dhaka, Bangladesh.