Acoustic Sensor for Voice with Embedded Physiology

The Army Research Laboratory (ARL) is developing sensor technology to monitor the soldier’s voice and physiology by using enhanced acoustic sensors. The physiological sensor consists of liquid or gel contained within a small, conformable, rubber bladder or pad that also includes a hydrophone. This enables the collection of high signal-to-noise ratio cardiac, respiratory, voice, and other physiological data. The pad also minimizes interference from ambient noise because it couples poorly with airborne noise. It is low cost and comfortable to wear for extended periods. When the sensor pad is in contact with a patient's thorax, neck, or temple region, sounds can be immediately and continuously monitored. This can aid in the remote assessment, diagnosis, and treatment of cardiac and respiratory functions, as well as provide human stress and performance indicators such as heart and breath rates, voice stress, and gross motor indicators. The neck sensor picks up the wearer's voice well, with fidelity sufficient to be used as a hands-free voice activation mechanism using speech recognition software, or for voice communications. The pulse is also detectable from the carotid artery, and excellent breath sounds are present as well. This attachment area is often unobstructed by other equipment or clothing, and is easy to attach quickly to a subject. Data presented will compare a microphone held in front of a speaker’s mouth to that of a fluid sensor held in contact with the neck. Comparing the amplitudes of the voice to the non-vocalized ambient noise surrounding the voice gives approximately 40 dB SNR for the airborne microphone, and approximately 75 dB SNR for the fluid-coupled sensor. The fluid coupling represents an improvement of better than 30 dB in SNR with minimal waveform degradation, as observed by the similar spectrograms and by listening to the data through headphones. The ability of body-coupled sensors to detect voice in high background noise was investigated. A gel sensor was attached to a speaker’s neck. Positioned in front of the person’s mouth was a boom microphone. Data presented shows simultaneously collected breath and voice data before, during, and after a speaking subject is submerged in a 105 dB C noise field while speaking a counting sequence repeatedly. The boom microphone did not detect any voice during the high amplitude noise. However, the counting is clearly visible in the spectrograms throughout the noise with the body-coupled gel sensor. Approved for public release; distribution is unlimited.