PECULIARITIES OF APPLICATION OF OBJECT RECOGNATION SYSTEM IN REAL TIME ON MICROCONTROLLERS WITH SUBSEQUENT VOISE OUTPUT OF INFORMATION FOR PEOPLE WITH VISUAL IMPAIRMENTS

Abstract

Purpose. The study of the minimum and maximum time required to complete one full cycle of object name recognition-announcement taking into account different word lengths, different object recognition speeds, as well as physical characteristics of visually impaired people for real-time object recognition systems on microcontrollers with subsequent voice output.

Methodology. Creating variants of combinations of words of different lengths, taking into account the possibility of setting the speed of speech generation in Espeak, and the average speed of speech in Ukraine. Calculation of the minimum and maximum distance to the object at the start of the recognition-announcement cycle. The minimum and maximum time required for a full cycle of object name recognition-announcement is set.

Findings. On the basis of the Espeak language synthesizer and the peculiarities of the Ukrainian language and speech, the time required to announce the names of objects of different lengths was investigated. The minimum and maximum time for completing the full cycle of information recognition-announcement is set, taking into account the physical characteristics of people with visual impairments, their speed of movement and the speed of reaction to voice information. The minimum and maximum distance to the object at the start of the cycle was also obtained, depending on the time required to complete one complete cycle.

Originality. The minimum and maximum time needed to complete the full cycle of information recognition and announcement was obtained, taking into account the physical characteristics of visually impaired people, the technical capabilities of modern neural networks and programs for speech synthesis, as well as the minimum and maximum distance to the object at the time of the start of the cycle. The minimum and maximum distance to the object at the start of the recognition-announcement cycle was studied.

Practical value. The obtained results can be used in the practical creation of online object recognition systems, to assess the possibility of using certain neural networks, based on the obtained minimum and maximum time for passing the complete cycle of recognition-announcement of information, as well as the time required for passing each of its separate elements.