This paper presents an innovative solution for electromyography (EMG)-based human motion analysis systems, addressing challenges of sensor comfort, inter-individual variations, and labor-intensive labeling processes. The solution combines textile towel-based electrodes with transfer learning techniques. The textile towel-based graphene/PEDOT:PSS composite electrode offers biocompatibility, low skin impedance, and user comfort, while transfer learning reduces the need for extensive new data labeling and enhances the generalisation ability of the motion analysis system. The proposed methodology achieves accurate classification of hand gestures with a minimal number of samples and epochs. This demonstrates the potential of transfer learning for efficient EMG-based human motion analysis.