Loughborough University
Browse

Interpretation of human breathing patterns towards a new augmentative and alternative communication solution

Download (6.84 MB)
thesis
posted on 2021-07-23, 09:35 authored by Yasmin Elsahar

Augmentative and alternative communication (AAC) encompasses a range of methods that replace or complement speech of individuals with complex communication needs. Predominant AAC methods rely on the interpretation of specific input signals, including purposeful hand gestures, eye-tracking vectors, and brain waves. However, most AAC systems still face limitations in terms of usage flexibility and cost, especially when a user’s speech impairment is compounded with a form of motor disability.

To address this shortfall, an alternative AAC solution, based on encoded modulated breathing patterns (MBPs), is presented to support the communication of specific categories of individuals with speech disabilities. This research focuses on the characterisation of human breathing airflow dynamics towards the establishment of a modulated breathing patterns interpretation (MBPI) approach supported by dynamic machine learning.

The research project aims to explore practical breathing patterns interpretation as an effective and user adapted AAC solution by utilising the characteristics of continuous modulated breathing patterns (MBPs) to output synthesised machine spoken words (SMSW). The physical mechanics of human breathing airflow are determined to characterise upper breathing airflow and identify the necessary mechanisms for the collection of the intended breathing signals.

A dynamic airflow pressure detection system (DAPDS) based upon the principles of breath characterisation is researched to detect MBP signals during the inhalation and/or exhalation phases, producing a set of distinct and recognisable patterns. MBPs are used to interactively encode words and phrases by assigning MBPs to user-selected SMSW, building up a lexicon of words and phrases. The MBPI platform utilises a dynamic time warping (DTW) with k-nearest neighbour (k-NN) supervised machine learning algorithm to recognise user incoming MBPs and output the corresponding SMSW for communication. Approved protocols testing 25 healthy subjects were designed to acquire user- selected MBPs. The results obtained express a mean reliability of 91.97% with increased user familiarity. The interpretation of breathing patterns sets the way for the creation of new breath decoding applications, including self-learning AAC systems that are able to respond to individual user needs.

Funding

Loughborough University

History

School

  • Mechanical, Electrical and Manufacturing Engineering

Publisher

Loughborough University

Rights holder

© Yasmin Elsahar

Publication date

2021

Notes

A thesis submitted in partial fulfilment of the requirements for the award of the degree of Doctor of Philosophy of Loughborough University.

Language

  • en

Supervisor(s)

Sijung Hu ; Kaddour Bouazza-Marouf

Qualification name

  • PhD

Qualification level

  • Doctoral

This submission includes a signed certificate in addition to the thesis file(s)

  • I have submitted a signed certificate