Real-Time Speech Recognition System for Robotic Control Applications Using an Ear-Microphone

摘要

This study is part of an ongoing research started in 2004 at the Naval Postgraduate School (NPS) investigating the development of a human- machine interface command-and-control package for controlling robotic units in operational environments. An ear microphone is used to collect the voice- activated commands providing hands-free control instructions in noisy environments Kurcan, 2006; Bulbuller, 2006. This study presents the hardware implementation of a theoretical Isolated Word Recognition (IWR) system designed in an earlier study. The recognizer uses a short-term energy and zero-crossing based detection scheme, and a discrete Hidden Markov model recognizer designed to recognize seven isolated words. Mel frequency cepstrum coefficients (MFCC) are used for discriminating features in the recognizer phase. The hardware system implemented uses commercial off-the-shelf (COTS) electronic components, in-ear microphone, is portable and costs under $50.00. The implemented speech capturing system uses the ear-microphone and the Si3000 Audio Codec to capture and sample speech clearly. The microprocessor processes the detected speech in real-time. The microprocessor s I/O devices work effectively with the audio codec and computer for sampling and training, without communication problems or data loss. The current implementation uses 1.181 msec to process each 15 msec data frame. Resulting recognition performances average around 73.72%.