Grant: Discovery Grants Program - Individual

New research avenues are explored in order to provide a reliable human-to-object interaction by using speech modality and radio frequency identification (RFID) technology. The aim is to integrate speech technology into RFID networks to allow human operators to communicate verbally with RFID devices, RFID-tagged objects and related information systems. The combination of speech and RFID technologies is expected to improve speed, accuracy, ergonomics and safety of numerous manufacturing and business operations such as item tracking and monitoring, warehouse management, and shipping. However, many problems related to the reliability of the speech component remain unsolved and constitute obstacles in the way of the deployment of speech-enabled RFID systems. To insure the efficacy of the proposed RFID-Speech framework in real-life conditions (manufactory and warehouse environment), a very robust real-time speech recognition and a non-constrained and flexible dialogue scheme should be provided. Therefore, our proposed strategy consists of operating at both acoustic and language levels. At the acoustic level, cutting edge algorithms based on bio-inspired approaches (bird flocking and genetic algorithms) are proposed to optimize the estimation and tracking of non-stationary noises. At the language level, the proposed strategy consists of leveraging knowledge in distributed ontologies to reduce confusability and ambiguity and to provide a reliable semantic confidence measure through a new concept of ontological distance. This concept is also incorporated into the dialogue manager component. These innovative solutions are expected to provide ubiquitous and intuitive access to the enterprise information systems by using speech/RFID modality, which will ultimately lead to important technological advances in real-life RFID applications of the manufacturing industry.