Smartphone-Based Mini-Program Voice Acquisition for Remote and Real-Time Acoustic Voice Analysis.

OBJECTIVE

We developed a WeChat mini-program called "Voice Acquisition" that can be installed on smartphones for remote and real-time acoustic analysis. The program analyzes multiple parameters, including the fundamental frequency (F0), maximum phonatory time (MPT), jitter, shimmer, noise-to-harmonic ratio, cepstral peak prominence, cepstral/spectral index of dysphonia (CSID), and mel-scale frequency cepstral coefficients. This study aimed to validate these acoustic measurements against the conventional Computerized Speech Lab (CSL) software and assess their potential utility for screening and follow-up of voice disorders.

METHODS

During the study, smartphones were used as the carrier for Voice Acquisition. Researchers simultaneously used CSL software and the Voice Acquisition program to collect and analyze the sustained vowel phonation /a/ from 77 patients with voice abnormalities and 80 healthy volunteers in both a soundproof room and a regular office environment. The correlation coefficients from both applications were calculated to establish the reliability of the mini-program. In addition, we compared the differences between volunteers and patients to explore their potential for screening and follow-up of voice-related disorders.

RESULTS

Voice Acquisition and CSL correlated moderately to highly strongly across most acoustic parameters (r = 0.543-0.989). MPT and F0 mean proved most stable: MPT showed near-perfect positive correlation across all conditions, while F0 mean ranked second, performing optimally in the soundproof booth. Conversely, F0 min exhibited the most dramatic context sensitivity, correlating strongly in the booth (r = 0.655) but becoming non-significant outside (P > 0.05). Absolute values differed slightly between systems. Some acoustic parameters (eg, CSID) from the mini-program significantly discriminated between pathological and normal voice groups.

CONCLUSIONS

Voice Acquisition provides acoustic measures that are strongly correlated with those from CSL, supporting its reliability for remote voice data collection. While preliminary data suggest potential for screening applications, further validation is needed to establish clinical utility.