@article{janke2017emg,
  title        = {EMG-to-Speech: Direct Generation of Speech From Facial Electromyographic Signals},
  author       = {Janke, Matthias and Diener, Lorenz},
  year         = 2017,
  month        = nov,
  day          = 23,
  journal      = {{TASLP} -- {IEEE/ACM} Transactions on Audio, Speech and Language Processing},
  volume       = 25,
  number       = 12,
  pages        = {2375--2385},
  doi          = {10.1109/TASLP.2017.2738568},
  abstract     = {Silent speech interfaces are systems that enable speech communication even when an
    acoustic signal is unavailable. Over the last years, public interest in such interfaces has
    intensified. They provide solutions for some of the challenges faced by today's speech-driven
    technologies, such as robustness to noise and usability for people with speech impediments. In
    this paper, we provide an overview over our silent speech interface. It is based on facial
    surface electromyography (EMG), which we use to record the electrical signals that control
    muscle contraction during speech production. These signals are then converted directly to an
    audible speech waveform, retaining important paralinguistic speech cues for information such as
    speaker identity and mood. This paper gives an overview over our state-of-the-art direct
    EMG-to-speech transformation system. This paper describes the characteristics of the speech EMG
    signal, introduces techniques for extracting relevant features, presents different EMG-to-speech
    mapping methods, and finally, presents an evaluation of the different methods for real-time
    capability and conversion quality.},
  url          = {https://halcy.de/cites/pdf/janke2017emg.pdf},
}