Evolutionary-Based Speaker Adaptation to Improve Interaction With RFID Networks in Noisy Environments

This paper presents an innovative application which consists of providing a speech modality to allow the human operators of a Radio Frequency IDentification (RFID) network to communicate verbally with related devices and information systems. Within this application, the results of two contributions are reported. The first contribution is a new framework for speech enhancement based on the Karhonen-Loeve Transform (KLT) optimized by Genetic Algorithms (GAs). The second contribution is an evolutionary-based approach for a fast speaker adaptation. The results of experiments using the ARPA-RM and NOIZEUS databases demonstrate the effectiveness of the proposed techniques.