Electrical and Computer Engineering ETDs

Publication Date

Fall 11-12-2020

Abstract

The focus of the research is to identify stress markers in a firefighter's speech. These markers include changes in breathing patterns and changes in the fundamental frequency of an individual’s voice. The breathing patterns are characterized using the number of breaths taken in a minute and the time spent inhaling. These measures are estimated using a Restricted Boltzmann Machine to process a firefighters’ SCBA regulator sounds, as open and closed. The classifications are then combined into continuous intervals. Observing the length of the intervals and the number of interval-starts represents time spent inhaling and the breathing rates (breaths per minute). The fundamental frequency estimation uses a set of Deep Long Short-Term Memory (LSTM) Recurrent Neural Networks (RNN). The first is used to classify speech segments as voiced/unvoiced. The fundamental frequency estimation is done on the segments using a deep LSTM RRN regression.

Keywords

Machine Learning, Nueral Networks, RNN, RBM, Health Monitoring, Signal Processing

Document Type

Dissertation

Language

English

Degree Name

Electrical Engineering

Level of Degree

Doctoral

Department Name

Electrical and Computer Engineering

First Committee Member (Chair)

Dr. Ramiro Jordan

Second Committee Member

Dr. Manel Martinez-Ramon

Third Committee Member

Dr. Balasubramaniam Santhanam

Fourth Committee Member

Dr. Edward G. Graham Jr

Fifth Committee Member

Dr. John Hansen, University of Texas Dallas

Download

Share

COinS