Breathing Pattern Interpretation as an Alternative and Effective Voice Communication Solution.

Area of Science:

• Biomedical Engineering
• Human-Computer Interaction
• Assistive Technology

Background:

• Current augmentative and alternative communication (AAC) systems often depend on gesture interpretation, which can be limited in versatility.
• Existing AAC methods may present challenges in terms of user-friendliness and adaptability.
• There is a need for more versatile and less cumbersome communication solutions for individuals with speech impairments.

Purpose of the Study:

• To investigate the feasibility of using breathing patterns (BPs) for communication.
• To develop a system for breathing-pattern interpretation (BPI) to encode messages with minimal user training.
• To explore BPI as a novel approach within augmentative and alternative communication (AAC).

Main Methods:

• Utilized a unidirectional microphone to capture analogue modulated breathing patterns.
• Applied single-channel Weiner filtering (WF) for signal de-noising of breathing patterns.
• Employed k-nearest neighbor (k-NN) classification with dynamic time warping (DTW) to interpret breathing patterns into synthesized machine-spoken words (SMSW).

Main Results:

• Achieved an 86% accuracy rate using k-NN classification for breathing patterns.
• Demonstrated decreasing error rates in live classifications for classes 2, 3, and 4 (17%, 14%, and 11% respectively).
• Reported a systematic reliability of 89% with increased user familiarity.

Conclusions:

• Breathing pattern interpretation (BPI) shows significant potential as a novel AAC method.
• The developed system offers a durable and potentially beneficial engineering solution for individuals requiring alternative communication.
• Minimal training is required, suggesting a user-friendly and accessible communication interface.