Advanced Machine Learning for Locomotion Intention Prediction

I am looking for someone to do a locomotion intention prediction project using wearable sensor data. The dataset contains .mat files from 10 participants performing 3 activities. It includes wearable sensor signals from different body locations, along with detailed transition labels. The goal is to predict the upcoming activity before the transition occurs. Because the dataset is limited, I am interested in advanced methods suitable for small datasets, such as contrastive learning, self-supervised learning, transfer learning, domain adaptation, or other suitable approaches. The freelancer should recommend the most appropriate method and evaluate it using subject-independent testing. The project should include data preprocessing, model, evaluation, and documented re runnable code in ipynb format.