SpeechOmeter: Heads-up Monitoring to Improve Speech Clarity
Mansoor Pervaiz, Rupal Patel · 2014 · Proceedings of the 16th International ACM SIGACCESS Conference on Computers & Accessibility (ASSETS '14) · doi:10.1145/2661334.2661339
Summary
This demonstration paper presents SpeechOmeter, a Google Glass application that provides real-time visual biofeedback on vocal loudness to help individuals with neuromotor speech disorders (dysarthria) speak more clearly during daily conversation. Dysarthria accompanies conditions such as Multiple Sclerosis, Parkinson Disease, and Cerebral Palsy, resulting in soft, breathy, fluctuating voice with imprecise articulation. Clinical speech therapy — particularly Lee Silverman Voice Treatment (LSVT) — teaches patients to increase vocal loudness, which has been shown to improve overall speech clarity. However, generalising these techniques from clinical settings to everyday life is limited by three factors: fewer than 5% of affected individuals have access to qualified clinicians, LSVT requires speaking at maximum effort which causes vocal fatigue, and patients need frequent reminders to maintain the strategies. SpeechOmeter addresses these barriers by providing unobtrusive, context-aware cues. The system consists of a mobile phone module that records ambient noise and a Google Glass module that records the user's voice, computes the difference in decibel levels, and provides visual feedback encouraging the user to speak at least 5 dB above the ambient noise — a relative target rather than an absolute one.
Key findings
In a longitudinal usability study, 12 individuals with Multiple Sclerosis increased their vocal loudness when provided with SpeechOmeter's visual feedback. The system's key innovation is measuring vocal loudness relative to ambient noise rather than using a static threshold, making it adaptive to different environments — a quiet room versus a noisy restaurant — without requiring manual adjustment. The 5 dB threshold was selected based on post-LSVT treatment data showing typical loudness increases in participants. The system also tracks usage and performance statistics that are shared with clinicians, enabling them to customise treatment plans based on real-world adherence data rather than relying solely on in-clinic observations. This clinician dashboard bridges the gap between therapy sessions and daily life, providing objective data on whether patients are applying their training outside the clinic.
Relevance
SpeechOmeter represents an early example of wearable assistive technology for speech disorders, addressing the critical "last mile" problem in speech therapy: transferring clinical gains to real-world conversation. For accessibility practitioners, the paper highlights that communication accessibility is not only about providing alternative modalities (like text or sign language) but also about supporting people in using their own speech more effectively. The relative loudness approach — measuring voice against ambient noise rather than against a fixed target — is a design principle applicable to many adaptive assistive systems. While Google Glass as a platform did not achieve widespread adoption, the underlying concept of heads-up biofeedback during natural conversation remains relevant to current smart glasses and wearable devices. The work is particularly pertinent for organisations supporting employees or students with neuromotor conditions.
Tags: speech disorders · dysarthria · speech therapy · wearable technology · biofeedback · multiple sclerosis · Parkinson's disease · cerebral palsy · assistive technology · Google Glass