ABS 010: Facial Muscle Activation During Speech: A Study of Language and Gender Differences Using sEMG ​

Wyatt DeBord ¹, Arina Kucinskaja ¹

¹ Georgia Institute of Technology

The Van Wickle Journal (2026) Volume 2, ABS010

Introduction: This study explored the relationship between facial muscle activation and spoken language using surface electromyography (sEMG). Despite widespread interest in the connection between speech and facial muscle usage, little research has examined how sEMG signaling varies across languages and genders. This research evaluated whether native language (English vs. Spanish) or gender (Native English Male vs. Native English Female) influences facial muscle activation during speech, with potential applications in linguistics, language learning, and speech pathology.

Two aims were addressed. Aim 1 assessed differences in sEMG activity between a sample size of 21 native English speakers and 21 native Spanish speakers. Aim 2 evaluated differences between 21 male and 21 female native English speakers. All participants (Georgia Tech students aged 18–25) read pangrams in their native language while sEMG data was recorded from the Orbicularis Oris, Depressor Angularis Oris, and forehead (Figure 1). Signals were filtered using a 20–400 Hz bandpass, the root mean square (RMS) amplitude was computed in MATLAB, and spectral density was analyzed. Two-tailed t-tests (α = 0.05) were used to assess group differences.

Results showed a significant difference in muscle activation between English and Spanish speakers (RMS p = 0.00005; spectral density p = 0.0124), but no significant difference between genders (RMS p = 0.2298; spectral density p = 0.1273). These findings suggest that language spoken influences facial muscle activation during speech, while gender does not. This supports language-specific targeting in speech rehabilitation and highlights the viability of standardized electrode placement across genders.

Methods: Participants consisted of 42 Georgia Tech students (21 E & 21 NE). Participants had electrodes placed on Orbicularis Oris (+), Depressor Angularis Oris (-), and forehead (GND). The E group was required to read a list of English Pangrams three times. The NE group read a Spanish Pangram list three times. The activation data was relayed back to an Arduino which compiled the raw data. Later it was filtered in Matlab with a bandpass of 20-400 Hz . Following this the root mean square (RMS) average of the amplitudes per participant were calculated and a two-tailed t-test was performed with an alpha level of 0.05. The second aim followed the same setup but with men and women instead of language.

Results: The results of the study indicated interesting new findings. For Aim 1, It was observed that there was no significant difference between genders speaking English. The p value for RMS was 0.2298 and the p value for spectral density was 0.1273, both of which were higher than the defined critical value, 0.05. However, the results for Aim 2 – English vs Spanish speakers - did indicate a significant difference. Here the p value for amplitude was 0.00005 and the p value associated with spectral density was 0.0124. Both were lower than the critical value of 0.05.

Discussion: These findings are particularly relevant to stroke rehabilitation studies as they indicate that while the gender of participants is irrelevant and unlikely to impact results, the native language of participants can impact the results. They also indicate that electrode-based treatments can be placed in the same muscle locations for male or female patients.​ They further suggest that tailoring treatments to native languages can improve results as Spanish and English activate the OO differently. Lastly, these results suggest that muscle activation can be utilized to differentiate spoken languages.