A new strategy for speech recognition through the parametric classification of ear pressure signals.

摘要

A new communication and control concept using speech is introduced to generate, detect, and classify signals that can be used in novel hands-free human-machine interface applications such as communicating with a computer and controlling devices. The signals are the changes in the airflow pressure that occur in the ear-canal caused by speech. The goal is to demonstrate that the ear pressure signals, acquired using a microphone inserted into the ear-canal, due to specific speech are distinct and that the signals can be detected and classified very accurately. The strategy for demonstrating the concept includes energy-based signal detection and segmentation to extract ear pressure signals due to speech, signal normalization to decrease the trial-to-trial variations in the signals, Welch's modified periodogram for power spectral density estimation, and band-divided averaging algorithm to estimate the spectral features from ensembles of pressure signals. The complete strategy of signal detection and segmentation, estimation, and classification is tested on 4 speech signals. Through extensive experiments, it is demonstrated that the ear pressure signals due to the speech are distinct and that the 4 pressure signals can be classified with over 92% classification accuracies using the Gaussian classification algorithm.