In this paper a new and original framework for handwriting to speech devices is investigated. This framework is based on a pen-shaped optical mouse connected to a human-machine interaction. The selected approach is divided in four steps: characters acquisition, characters digitalization, characters recognition and speech generation. Characters acquisition uses a pen-shaped optical mouse connected to a USB port. Characters digitalization is based on the extraction of the coordinates of the pen positions during the drawing of the characters, which are then sent to the character recognition module. This module is based on a artificial neural network algorithm. The identified letters are then put together to build words and possible typesetting errors are corrected. These words are sent one after the other to the speech generator to be pronounced which allow disabled people to communicate easily. Compared to others approaches, the key points and advantage of the proposed framework are the robustness of the algorithm, the real-time interaction, the size and weight of the device (and the price).