Unstable Morse Code recognition system with back propagation neural network for person with disabilities

Area of Science:

• * Assistive Technology
• * Machine Learning
• * Signal Processing

Background:

• * Traditional Morse code auto-recognition systems struggle with unstable typing speeds and interval ratios.
• * Existing algorithms are insufficient for applications requiring recognition of non-standard Morse code patterns.
• * Unstable Morse code time series present significant analytical challenges for current methods.

Purpose of the Study:

• * To develop and evaluate a neural network-based system for recognizing unstable Morse code patterns.
• * To assess the system's performance with users exhibiting significant typing variability.
• * To demonstrate the feasibility of real-time Morse code recognition using neural networks.

Main Methods:

• * Implementation of a neural network architecture designed to analyze time series data.
• * Training and testing the neural network on Morse code input from individuals with typing impairments.
• * Comparative analysis of recognition rates across different user groups (cerebral palsy, amputee, expert).

Main Results:

• * The neural network achieved an average recognition rate of 93.2% for a teenager with cerebral palsy.
• * An average recognition rate of 97.2% was recorded for a 40-year-old amputee using a prosthesis.
• * The system demonstrated high performance, with results comparable to a skilled expert's 99.2% recognition rate.

Conclusions:

• * The neural network effectively overcomes the challenges of analyzing severely unstable Morse code time series.
• * The system shows significant potential for improving communication accessibility for individuals with typing disabilities.
• * The computational efficiency allows for potential real-time signal recognition applications.