Advancing Electromyographic Continuous Speech Recognition: Signal Preprocessing and ModelingSpeech is the natural medium of human communication, but audible speech can be overheard by bystanders and excludes speech-disabled people. This work presents a speech recognizer based on surface electromyography, where electric potentials of the facial muscles are captured by surface electrodes, allowing speech to be processed nonacoustically. A system which was state-of-the-art at the beginning of this book is substantially improved in terms of accuracy, flexibility, and robustness. |
Other editions - View all
Advancing Electromyographic Continuous Speech Recognition: Signal ... Michael Wand No preview available - 2020 |
Common terms and phrases
acoustic data action potential adaptation data amount of training application articulation articulatory audible and silent audible EMG Average Word Error background system Bars indicate standard baseline system BDPF models BDPF streams chapter computed context width context-dependent corpora cross-mode decision tree decoding vocabulary described in section dimensionality reduction electrode arrays Electromyography EMG arrays EMG channels EMG data EMG signal EMG-based speech recognition EMG-PIT corpus EMG-PIT pilot corpus EMG-UKA corpus entropy entropy gain evaluation experiments feature extraction Figure frequency Independent Component Analysis indicate standard deviation input manner of articulation method MLLR motoneuron multi-mode systems muscle node observed optimal parameters performed recognizer session-independent systems silent EMG silent speaker Silent Speech Interface speaker speaking mode Spectral Mapping speech sounds thesis tion training data training sentences unit models utterances Viterbi algorithm vocal tract vowel whispered EMG whispered speech Word Error Rates yields