Electromyography based Word Recognition for Silent Speech Interface
One of the electronic systems that enable to communicate by speech without an audible acoustic signal is Silent Speech Interface. The work investigates the usability of muscle contractions detected by surface electromyography (EMG) sensors as an input channel for Silent Speech Interface. Therefore, the technology enables speech recognition to be applied to silently mouthed speech. The work describes the outcomes in using artificial neural network to recognize and classify human speech based on EMG signals which are captured at the facial muscles, records the activity of the human articulatory apparatus and thus allows tracing back a speech signal even if it is spoken silently. Since speech is captured before it gets airborne, the resulting signal is not masked by ambient noise. The preliminary results demonstrate that the proposed technique yields high recognition rate for classification of unvoiced words using SEMG features. The output of Silent Speech Interface has the potential to overcome major limitations of conventional speech driven interfaces: it is not prone to any environmental noise, allows to silently transmitting the confidential information, and does not disturb bystanders. The results demonstrate that the system is easy to train for a new user. This work forms the basis for further researches to use EMG signals to improve large training dataset and uses of different languages based model.