Bachelor Graduation Project Tracheo-esophageal speech enhancement

Abstract

Every year an unfortunate part of the population lose their larynx, often due to the consequences of cancer. The laryngectomy removes the possibility to speak. Fortunately several speech reinstatement techniques have been found. However, all current speech reinstatement techniques do not produce a lot of volume. As a result the laryngectomised person will have difficulties speaking in busy environments, up to the point were the condition can lead to isolation because of the inability to communicate. Moreover with the current methods it costs a lot of energy to speak, the produced speech is less stable and the fundamental frequency has become lower. In this thesis a device is designed to improve the speech produced by a laryngectomised person realtime. The speech is recorded with an electret microphone. The resulting signal is prepared to be quantified to then be processed by a digital signal processing unit. Then the signal is adaptively filtered using the dynamically determined fundamental frequency. By tracking the fundamental frequency some instability from the voice is removed. The fundamental frequency will be determined using the cepstrum, a technique separating the fundamental regency of a voice from the high frequency components added by the oral and nasal cavities during speech. In another thesis a system is designed to amplify the signal to make sure that speaking takes less energy \cite{shiftgroep}. Furthermore a suitable power source is determined for the system. As the system is designed for Tracheo-Esophageal Speech Signal Amplification, the system will be referred to as TESSA.