Speaking a foreign language without an accent? Mouth sensors should make it possible

Speaking a foreign language without an accent could soon be easier thanks to mouth sensors. Scientists at TU Dresden plan to develop a novel system for computer-based pronunciation training over the next three years in the project "ADAMA - Accent Improvement through Pronunciation Training with Articulatory Feedback".

Learning to speak a foreign language without an accent is a great challenge, especially for adults. A completely new movement pattern for the speech organs must be learned. The muscles of the palate, tongue and lips must adapt to the new sound formation. This requires regular and correct practice. The learner also needs the opportunity to receive comprehensive feedback. Therefore, scientists at TU Dresden are working on the development of a virtual coach that takes over this task. "Previously available computer systems for learning a foreign language only evaluate pronunciation on an acoustic basis and do not support the improvement of speech quality," says project collaborator Alexander Wilbrandt from the Junior Professorship for Cognitive Systems at TU Dresden. "They do present the user with the need to improve his pronunciation, but do not show where the pronunciation errors come from and how he or she can fix them. We, on the other hand, want to develop a training program that offers concrete advice on correct phonation and can be used for targeted accent reduction."

In the ADAMA system, in addition to acoustic measurements, which are performed with a microphone, articulatory measurements are also realized by a sensor system placed directly in the mouth. The sensors are installed on a flexible circuit board that can be easily inserted and removed. This allows a novel acoustic-articulatory matching procedure to be performed. In an interactive learning environment, the trainee receives feedback on the acoustic-articulatory deviation from the optimal native reference. Visual feedback shows both the current and target positions of the tongue, jaw, and lips. "If the foreign language target sound is presented purely acoustically, i.e., without visual feedback, the learning effect is lost, because the learner will try to produce the unfamiliar sound using speech movements familiar to him or her in the native language," Wilbrandt explains.

"ADAMA" is funded by the German Federal Ministry of Education and Research (BMBF) with nearly 350,000 euros. At the end of the project, a functional demonstrator of the overall system is to be created, which will then be prepared by project partner Linguwerk GmbH for market entry during 2024.

Information for journalists:
Alexander Wilbrandt
Tel.: 0351 463-33106