Abstract

Nowadays, it is becoming increasingly affordable to enhance the home environment with several automation schemes, allowing remote control of e.g., heating/cooling, communication, lighting, media, etc. Such smart home functionalities are essential for people with disabilities and the elderly, as they not only provide assistive control of important everyday functionalities, but may prove to be life-saving in case of emergency. However, smart home functionalities may become useless for people who need those most, if they cannot be accessed via a natural, easy to use, interface. The central objective of LISTEN is to design and implement a complete system, including both the software and hardware components, enabling robust hands-free large-vocabulary voice-based access to Internet applications in smart homes. This would allow the users to have natural control (i.e., using their voice) of the smart-home web-enabled functionalities (e.g., turning on/off web-enabled “smart” appliances), but also to access specific Internet applications (e.g., web search, email dictation, access to social networks). A truly hands-free system operation of the voice interface is equally important: users will not have to turn towards a microphone or other device, or wear a headset. Therefore, LISTEN will develop (a) a robust hands-free speech capture system operating as a wireless acoustic sensor network (WASN), specifically designed for the smart home, and (b) a large-vocabulary automatic speech recognition system optimised for accessing web applications and controlling web-enabled smart home automation functionalities. LISTEN pushes the boundaries of current state-of-the-art by bridging the gap between the acoustic front-end and automatic speech recognition research communities, with the common goal of developing a smart-home-specific natural voice interface to web services.