ForkLocator

Locator band prototype:

Testing movie clip:

The hardware design is straightforward. The flappers will be installed on the motors and pressed on the user's arm gently and repeatedly to indicate the direction to move. 
The device will be separated into the following modules:

1. left flapper
2. right flapper
3. forward and backward flapper
4. controller  module

All the modules would be installed onto the elastic band eventually to make the device portable and easy to use.

 As this is still the POC stage,  the battery is not included and the wirings would be handled afterward.

For the left/ right flapper module:

For the forward/ backward flapper module:

For the Arduino and Bluetooth module:

Overview:

Prototype Requirement:

1. To identify the fork and arm (hand)
2. To show existence of the fork and arm
3. To indicate in which direction the arm should move in text

Object Detection:

Yolo has been chosen for object detection. In comparison with CNN or typical machine learning method, Yolo would give better performance in terms of speed, accuracy and flexibility. 

The following is the learning and testing results:

The next step would be the hardware side

Considerations:

1. able to locate the arm(hand), fork and plate on the table
2. able to give a clear notification to the visually impaired
3. the outcome would be a portable device 
4. (optional) Others would not know while using the device

Requirement:

1. object identification <1s
2. the actuator should be as small as possible
3. the device is wearable

The process flow:

1. The mobile camera would be used for object detection.
2. After the objects were recognized, the corresponding actuators would be activated to indicate the arm movement.
3. When the arm and target are very near, there would be another signal to indicate the user to pick the target.