Embedded Deep Learning to Aid the Mobility of Individuals with Disabilities: A Solution for In-house Bus Line Recognition
Date Issued
2022
Author(s)
Pavlov, Tashko
Stanojkovski, Nikola
Stojchevski, Mario
Simjanoska, Monika
Abstract
The mobility of individuals with visual impairments is a significant challenge as the cities are becoming more and more crowded each
day. The technology is rapidly developing, offering novel high-tech smart
white canes to aid the mobility of individuals with partial or total blindness. However, they are hardly affordable due to the high prices. Even
more, they are impractical for in-vivo usage as they depend on thirdparty technologies and services, which require an Internet connection for
data transfer and data processing on Cloud services. In this paper, we
offer a novel methodology that aids the transportation of blind individuals, which is entirely integrated into the chip, thus avoiding the need for
an Internet connection. Our methodology embeds three intelligent Deep
learning models on a single smart mobile device, one model to localize the
position of the bus line number approaching the individual, the second
model to recognize the bus number, and the third is a text-to-speech
model, which synthesizes speech to notify the individual in a pleasant
and human-like manner about the number of the approaching bus. Our
work presents one step closer to the completely independent embedded
intelligent models that simplify the transportation of visually impaired
persons using cutting-edge tools from AI.
day. The technology is rapidly developing, offering novel high-tech smart
white canes to aid the mobility of individuals with partial or total blindness. However, they are hardly affordable due to the high prices. Even
more, they are impractical for in-vivo usage as they depend on thirdparty technologies and services, which require an Internet connection for
data transfer and data processing on Cloud services. In this paper, we
offer a novel methodology that aids the transportation of blind individuals, which is entirely integrated into the chip, thus avoiding the need for
an Internet connection. Our methodology embeds three intelligent Deep
learning models on a single smart mobile device, one model to localize the
position of the bus line number approaching the individual, the second
model to recognize the bus number, and the third is a text-to-speech
model, which synthesizes speech to notify the individual in a pleasant
and human-like manner about the number of the approaching bus. Our
work presents one step closer to the completely independent embedded
intelligent models that simplify the transportation of visually impaired
persons using cutting-edge tools from AI.
Subjects
File(s)![Thumbnail Image]()
Loading...
Name
embedded-deep-learning-to-aid-the-mobility-of-individuals-with-disabilities-a-solution-for-in-house-bus-line-recognition.pdf
Size
2.03 MB
Format
Adobe PDF
Checksum
(MD5):5144f7bd79b2bc4430cfc2c3b977c0d5