Smart glove for speech impaired individuals

Features

1. Flex sensor based detection of signs and custom android app for text and audio output (using MIT app inventor)
2. Employs fingerspelling, a new mode of communication employing 3 modes and entire device is operated with signs, including START and STOP a. 3 Modes each having 32 options to select b. Mode 1 - common phrases c. Mode 2 - alphabets d. Mode 3 - Numbers and special characters e. Start and Stop
3. Custom created dataset of ADC values from 100 individuals - total 88000 set of values, each class 1000 values
4. ML algorithms embedded to increase the accuracy from 79% to 91%

References

1. Report submitted for Final year project
2. Image of signs
3. Dataset analysis and ML application on the dataset
4. Code for quickly finding which ML models to use among many of the ML models using LAZY PREDICT library

Codes

1. Data collection from arduino (ADC values)
2. Sending the output to app
3. Read data from arduino
4. Flex sensor ADC value to decimal number convertor
5. Main loop

Files that can be given only after request for duplication or advancement of the project

1. Dataset file for training and testing ML models
2. Models file