University of Washington researchers have developed a new “spatial speech translation” system that allows headphone users to translate multiple speakers simultaneously while still preserving each speaker’s voice direction and quality. This emerging technology helps address the limitations of existing translation tools, which often only can focus on one single, isolated speaker.
On a visit to a noisy museum in Mexico, lead author of the study Tuochao Chen found it very difficult to use translation apps. This motivated him to develop the system, which uses artificial intelligence to help detect speakers in chaotic physical spaces and track them wherever they may move, simultaneously translating their speech and delivering it through noise-cancelling headphones. Chen’s team’s spatial speech translation system currently supports Spanish, German and French, but could expand to around 100 languages in the future.