A distributed speaker identification system for personalized home control

This paper presents a combined speaker identification/speech recognition system that can be efficiently used for personalized home control. The proposed system works as a distributed framework and is designed to identify a speaker in home environments in order to provide user access to customized options. Human speech signals contain both language and speaker dependent information. Using this information the system realizes a personalized control in home environment and this approach can also be applied in more generic scenarios such as car customization settings. The system was optimized with the aim to allow an immediate use only with the addition of small and cheap audio front-ends that will capture his spoken commands. Meanwhile a remote server performs the speech recognition as well as user identification and combines these informations to provides user specific settings which are sent back to the desired actuator at home.