Voice Trainer

Designers: Calvin Lee, Joe Hoerner, Morgan Fox

Client Coordinator: Meredith Nye, SLP, CCC, Duke University Medical Center

Supervising Professors: Kevin Caves and Richard Goldberg


People with Parkinson’s disease (PD) often develop of speech and voice disorders.  Our client, an older woman with PD, experiences these common symptoms as she is prone to gradually lowering the volume of her voice and has little vocal endurance. Many patients with compromised speech secondary to PD, including our client, undergo the Lee Silverman Voice Treatment (LSVT), a method of therapy that seeks to improve speech and voice by focusing on improving vocal loudness.

In LSVT therapy sessions, patients are reminded to speak more loudly when their voice drops below an appropriate volume, and they also undergo various vocal loudness exercises. LSVT has been shown in multiple studies to be an effective program in treating speech and voice disorders.Therapy is intense, requiring patients to commit to fourconsecutive weeks of daily therapy sessions intended to habituate speaking loudly.  We developed the voice trainer to support LSVT therapy by providing biofeedback on voice volume throughout the day.


Our client’s therapist reported: “The device will be helpful in giving patients biofeedback to help ensure they are speaking with sufficient volume.  It often falls to the user’s caregiver or spouse to remind the user to speak up and it is not uncommon for this to lead to resentment. This device will definitely help improve the user’s quality of life.”


The team developed a microprocessor-baseddevice that measures vocal and ambient volumes and gives feedback if the user is not speaking loudly enough. It can be worn or carried and monitors speaking volume continuously. If the speaking volume is not loud enough compared to the ambient volume, a red LED is illuminated and a vibration motor is activated.  If the speech is loud enough, a green LED is illuminated and there is no vibrational feedback.

The device uses an omnidirectional microphone to measure the speaking and ambient volume and a throat microphone to detect whether or not the user is speaking.  The throat microphone used in this device is a Socom Paintball Throat Microphone (RAP4 #001981). It has an adjustable strap with a magnetic connector that provides a snug but comfortable fit.  The microphone works by sensing the user’s vocal vibrations, so it does not pick up sounds from other peoples’ speech or ambient noise. Although signal magnitude increases with speech volume, this microphone cannot be used to reliably measure volume because signal magnitude is very sensitive to positioning and tightness of the strap.  The omnidirectional microphone was taken from a Communications Ultra-Light Headset (Radioshack #19-315).  Because the microphone is omnidirectional, its signal is not sensitive to positioning. This microphone is attached to the throat microphone strap, rather than on the user’s clothing, in order to ensure consistent microphone placement and to consolidate wires (Figure 1).

Both microphone signals are input to a custom circuit, which filters and amplifies the signals and passes them to an ArduinoMini microcontroller.  This microcontroller was chosen because it is easy to program, yet powerful enough to rapidly sample and process the analoginputs. It samplesthe throat and omnidirectional microphones signals, determines whether speech is sufficiently loud, and gives an appropriate output.

A small, ABS-plastic case was fabricated and the device can be kept in a pocket or clipped onto an article of clothing (Figure 1). The case contains the PCB, microcontroller, rechargeable 7.4V, 800mAh Polymer Lithium-ion battery, on/off switch, and feedback mechanisms. It is designed so that feedback can be easily noticed by the user; the vibration motor is positioned to maximize sensation and the LEDs are placed in a convenient location.

The total cost of the device was $232.

Comments are closed.