Exploring the Feasibility of Using Smartphones for Measuring Sound-Level Difference as a Treatment Outcome.

Abstract

Purpose: In the last decade, acoustic voice analysis using smartphones and freely downloadable apps has gained popularity. This observational study aimed to establish the feasibility of using uncalibrated smartphone devices with internal microphones as sound-level meters (SLMs) of voice in suboptimal clinical environments with ambient noise for sound-level difference measurement.

Method: Four smartphones (two iPhones and two Androids) were tested against a Class 2 Casella SLM. Within-device consistency and linearity were tested across a range of synthesized tones and recorded sustained /a/ vowels, adjusting for app choice, microphone-to-source distance, sound type, intensity, frequency, and ambient noise level (< 30, < 50 dB). The strength of within-device test-retest reliability was also assessed.

Results: Across devices, single dB values differed widely. Under all conditions, a strong linear relationship was seen across the Class 2 SLM and all smartphones (R² > .980). An increase in ambient noise from < 30 to < 50 dB did not affect the linearity of any of the tested devices (R² > .980). The test-retest reliability showed near-perfect linearity across measurement iterations and between devices, R² > .998, τ > .99. Measured differences between white noise tokens across all devices showed good agreement, intraclass correlation coefficient (2,1) = .88.

Conclusions: Provided that internal linearity can be proven, uncalibrated smartphones demonstrate a feasible method of measuring pre-/posttreatment sound levels in sustained vowels provided that the device microphone-to-source distance and recording parameters (i.e., time and frequency weighting and application) are the same. Conversely, for absolute sound pressure level measurement, calibration with a gold standard SLM is needed.