Methods, systems and devices are provided for controlling a speech recognition process on a computing device. A computing device may receive audio signals from a microphone and muscle movement signals from a muscle movement detector, such as an electromyography sensor. The computing device may determine whether the audio signals satisfy an audio characteristic criterion indicative of speech and whether the head muscle activity signals satisfy a muscle movement criterion indicative of the user speaking. The computing device may perform voice recognition processing on the audio signals through a voice recognition algorithm in response to determining that both the audio signals and the head muscle activity signals satisfy their respective criterion indicative of the user speaking, and not perform voice recognition processing of audio signals while either audio signals or head muscle activity signals do not satisfy their respective criterion indicative of the user speaking.