Robust recognition of linked figures with limited resources

The wireless telephone often imposes significant restrictions on the size and cost of components of the transmission system because the size and cost of the materials are important factors for the development of a product or service. Until now, few products in cellular telephony offered services such as automatic speech recognition. The most immediate need is for hands-free and eye-free communication, since many cellular users wish to place calls without concentrating on the keyboard. The possibilities of service often pass the recognition of connected figures for the access of database information. Our objective is the realisation of a system of automatic recognition of a limited vocabulary using little memory and calculation for applications in cellular telephones, including hands free systems. A simple procedure founded on the segmentation of syllables was developed for the automatic recognition of linked figures utilising the base data of TIDIGITS.