In this study, we implemented off-line wearable CNN based finger number gesture recognition system on Raspberry pi. OpenBCI’s bio-sensing board was used to collect sEMG training signals and sEMG test signals collected from the forearm of a subject when finger number gesture is performed. Using training data, The CNN model was trained on the Google Colaboratory platform, and the resulting model was deployed and executed on the Raspberry Pi board. In experimental study, CNN model on Raspberry pi recognizes finger numbers from one to five with an accuracy of 91.3%.