While smartphone users have been able to translate spoken language into almost any modern language for many years ago, this feat was incredibly difficult to achieve using small microcontrollers that are memory-limited. In response to this problem, Hackster.io user Enzo chose to develop an experiment that proved the way it is possible for an embedded device can identify the language in use without the requirement for the Internet connection.
The”language detector” or “language detector” is based on an Arduino Nano 33 BLE Sense that is connected to a PCA9685 common motor driver, which is connected to three micro servoswhich are all powered by one 9V battery. Enzo made a data set with three phrases: “oui” (French), “si” (Italian)”si” (Italian) and “yes” (English) for about 10 minutes each, making the total of 30 minutes of audio files. The researcher also recorded three minutes of background noise that helps distinguish between the key words he was targeting and other words that are not important.
Once a model has been developed by using Edge Impulse, Enzo exported it to the Nano 33 BLE Sense and added a tiny piece of code that listens to audio coming from the microphone, categorizes it, and decides on which language is being spoken. Based on the results that is determined, the flag of the country is raised to signify the country’s language.
The post This Arduino device can detect which language is being spoken using tinyML appeared first on Arduino Blog.