Deep Learning Frameworks for EMG-Based Hand Movement Forecasting Internship

About Internship

Assistant Professor from Indian Institute of Technology, Patna (IIT Patna), is accepting interns interested in the field of Computer Science Engineering, in the research area of Artificial intelligence and Machine learning.

The prospective candidate will be responsible for the end-to-end development of a deep learning framework for forecasting natural and intuitive hand movements from electromyography (EMG) signals. The role will begin with the systematic handling of data, synchronization, and organization of multi-channel EMG time-series datasets. A key responsibility will be the implementation of robust preprocessing pipelines, which involve signal denoising (using band-pass and notch filtering), normalization, segmentation using sliding windows, resampling as needed, and handling of missing or noisy channels to ensure data quality and reproducibility. The candidate will explore both handcrafted and learned representations of EMG signals. Feature extraction, where applicable, will include time-domain, frequency-domain, and time–frequency features, such as RMS, MAV, WL, zero-crossing rate, spectral entropy, and wavelet-based descriptors. This will be followed by appropriate feature scaling and dimensionality reduction, as needed. Model development will involve designing, training, and benchmarking multiple deep learning architectures, including LSTM and CNN–LSTM models for temporal–spatial dependency learning, transformer-based models for long-range temporal attention, and hybrid CNN–Transformer frameworks to jointly capture local signal patterns and global temporal dynamics. Hyperparameter tuning, cross-validation, and performance evaluation using appropriate regression and forecasting metrics will be integral tasks. The candidate will also be responsible for thorough documentation, maintaining clean code repositories, preparing technical reports and manuscripts to ensure transparent, reproducible, and publication-quality outcomes. Preference will be given to candidates who have already published or have accepted a couple of relevant research articles in peer-reviewed venues.