Russian Academy of Sciences St. Petersburg Federal Research Center (SPb FRC RAS) have developed a mobile application that reads the user’s speech from their lips in noisy places. It has been reported TASS with reference to the press service of the scientific institution.
“Researchers have learned to recognize human speech from the lips using artificial intelligence algorithms and computer vision,” the press service said.
According to the representative, the enhancement will help improve the accuracy of voice assistants in crowded places or when using heavy equipment.
In their development, the Russian Academy of Sciences St. Petersburg Federal Research Center used a dual method of data collection, analogous to human perception – the application collects both the sound form of speech and lip movements.
“According to the scientists, the created neural network can automatically decide which data, when recognized, will provide maximum accuracy,” the press service said.
During testing of the application by noisy truck drivers from one of the logistics companies in Russia, the recognition accuracy exceeded 90%.
It was previously reported for the Main Radio Frequency Center (GRC) affiliated with Roskomnadzor. develop a service based on artificial intelligence that will allow you to check video recordings of conversations for lies and manipulations.