Helping Hand - AI Powered Glove to Assist the Blind

• 

• 

  Helping Hand Prototype Glove

Inspiration

Visually impaired/blind people need to memorize the location of all objects in order to navigate and pick up objects. In a completely new environment blind people aren't able to find the things they need or to avoid obstacles. This not only challenging for the blind to adapt to a new environment naturally, but can also induce a sense of helplessness. Our goal was to increase the quality of life for the blind and give them a tool to help them navigate new environments in life.

What it does

In order to combat this issue, we created the Helping Hand glove that assists blind people in finding objects based on sensory feedback from a camera and processed with a machine learning pipeline (powered by AlwaysAI). The webcam is able to detect over 1000 common everyday objects, lock on to the target, and guide the hand toward the object. While this is the prototype's primary use, this can be adapted into other wearables such as glasses and necklaces with integrated cameras that can notify and/or assist the user in everyday life.

How We built it

We used a Raspberry Pi 4 to run the machine learning model that classifies and tracks objects. We also used an Arduino for running a buzzer that notifies the user to move their hand left or right. This currently runs off of a 5v 3a power supply, but future iteration can be adapted with low-power compute devices and a portable battery.

Challenges We I ran into

Challenges we ran into was setting up the computer vision/machine learning model. That was before we attended the AlwaysAI workshops which simplified the entire process of integrating computer vision. Other challenges included finding a use case for the project as well as sorting out some minor bugs.

Accomplishments that I'm proud of

We are proud of the fact that we were able to accomplish so much in our first hackathon, especially as an all freshman team with little to no experience in electrical engineering and software engineering.

What We learned

We learned that 24 hours is not enough time to polish everything, but we worked through many issues and design iterations and ideas throughout the whole hackathon.

What's next for Helping Hand

We would like to reduce the size of the computer vision system enough to the point where it is feasible enough to actually manufacture this product to benefit blind people.

Built With

• alwaysai
• arduino
• electrical-engineering
• logitech
• machine-learning
• models
• neural-nets
• python
• raspberry-pi
• webcam