This project was completed alongside Smaran Velidi and Aditya Vishnubhotla.
Those with limited eyesight or complete blindness face various problems, including risks of physical injury, restrictions on daily activities, and negative impact on learning or education.
The white cane is the most common solution, as it is cheap and simple to make. Unfortunately, the white cane (and alternatives like guide dogs) have various problems - they are limited in scope and inefficient to use. The white cane also makes one hand unusable, whereas a guide dog can be extremely expensive.
Most of all, those using visual aids like the white cane or guide dog are subject to social stigma. In this project, my group and I sought to create an integrated/hidden alternative to visual detection with high accuracy and precision.
This project had the following engineering goals:
- Develop an object-detection device that can be both accurate and efficient at detecting nearby objects and notifying the user of their presence
- Restore normal functionality to both hands via a less intrusive and automatically functional device
- Create a cheaper visual-aid option for the visually impaired
To reach these goals, our group decided to create the BioProtect system.
The BioProtect vest reads in the surroundings by finding local objects via a phased array of ultrasound sensors and the vector intersections (indicating an object). Then, the user is alerted via the inflation of a pressurized pouch in the relevant direction. For example, if there is an object to the right of the user at mid-chest level, the pressurized pouch on the right size of the user at mid-chest level is to inflate.
Below is a schematic of the BioProject vest:
Here is an example of the field usage of the BioProtect system:
Before testing the device, several optimizations were done to the device. The most important optimization to the device were the addition of the pressurized compartments mentioned above. We initially had a verbal system (for example, a speaker would relay to the user the direction and distance of local objects) similar to most other prototypes [1]. However, adding the pressurized compartments proved to be an invaluable decision, as the new system was far more intuitive and less time-sensitive (the user was aware of local objects immediately rather than having to wait for an audio cue).
Another obstacle to be fixed was accuracy - our initial models either required bulky devices (such as interfacing with an Xbox Kinect) or were too inaccurate to be used successfully. To solve this, we decided upon using a phased array of ultrasound sensors, as described below:
The next optimization, battery interrupts, is described below. This optimization was important to ensure that the device was useable all day long.
Another issue to overcome was biomechanical adaptation - many existing prototypes were unable to cope with changes to the user's gait (which occurs when tired, after injury, due to normal growth during aging, etc.). Our final optimization, as described below, was total biomechanical adaptation:
Two primary modes of data collection and data analysis were mapped - simulated and experimental (physical). To verify the model's feasibility, we first created a CAD model and tested the device accuracy. After those tests were completed, we created a prototype physical model.
Below are the results for the Simulated and Experimental phases, with the data analysis in the subtitles:
Simulated Results
Experimental Results
We also compared BioProtect to an alternative prototype (found here) to determine BioProtect's comparative efficacy:
From this project, we reached the following conclusions:
- The Ultrasonic sensors, Kinect sensors, and Pressure pockets were inserted within the design of the vest and custom-fitted to minimize protrusion
- The real-life trials indicated that there was ~97% accuracy in randomly generated scenarios
and 95% in complex (testing scenarios)
- The biomechanical adaptation system mitigated errors by adapting to the specific gait of the user
As a result, the BioProtect vest succeeded in its engineering goal and is a feasible visual aid replacement.
The following future research could be conducted to further improve the system:
- A visual-analysis sensor could be added to the vest, to introduce braille translation capabilities
- Incorporating other sensors to increase overall accuracy and field of operation
(Arduino Ultrasonic HC-R04)
- Could evaluate the viability of the technology in a military setting, to detect threats in
object-dense areas with limited vision
If you'd like, you can view the extended presentation of the project (used at scientific conferences/competitions) here. While this presentation has a more thorough overview, it was conducted prior to experimental testing and does not contain a large portion of the final data.
Thanks For Reading! If you have any questions, or would like to get in contact, you can reach out to me at aruneswara@icloud.com.