Abstract: Gestures are not only an efficient source of human-computer interaction information but also an effective means of communication for individuals with hearing and speech impairments. Traditional gesture recognition, which relies on non-contact images or video information, suffers from issues such as light interference and the complexity of multidimensional signal analysis, leading to low accuracy and efficiency. Capturing the electromyography (EMG) signals caused by hand gestures through flexible skin electrodes is an effective way to overcome these limitations. This study proposes a conductive polymer flexible skin electrode made by mixing Multi-Walled Carbon Nanotubes (MWCNT) and Ecoflex polymer. When the doping level of MWCNT in Ecoflex reaches 8 wt%, the relative resistance change of the composite film is only 42% under 50% tensile strain; and it shows high stability in 10,000 cycles of stretching-releasing tests. In surface electromyography (sEMG) measurements, this skin electrode demonstrates lower noise than traditional commercial Ag/AgCl electrodes and is reusable multiple times. By attaching this skin electrode to the surface of the flexor carpi radialis muscle on the forearm, it captures the sEMG signals caused by hand gestures. Combined with convolutional neural network algorithms, it trains and tests six types of gesture signals, achieving an average accuracy rate as high as 98.5%. This research provides a more natural and portable means of communication for individuals with hearing and speech disabilities, and offers significant tools and technical support for fields such as human-computer interaction and medical rehabilitation.