The application of classification techniques to the heart rates of miners in escapeways

The techniques of signal processing and pattern classification have been applied, in this thesis, for the classification of the miners' heart rates during rescue operations. These techniques were used because of the limitations noted of the traditional statistical approach used in analyzing the nonstationary heart rates of the miners during rescue operations. The properties and application scope of wavelet are discussed. An appropriate wavelet, namely db4, is selected to perform a dyadic decomposition of the miners' heart rates. The decomposition results are prepared for extracting the feature vectors. Some programs developed in MATLAB language are used for extracting feature vectors from the decomposition results of the miners' heart rate data. Pattern classification techniques are discussed. One program package of classification, based on the LBG algorithm, is applied for the classification of the miners' heart rates. The classification method used produced encouraging result in spite of limited sample size. The result of the study provides a significant trial in analyzing the nonstationary time series, such as the heart rates of miners during rescue operations, by using signal processing and pattern classification techniques. A total of 37 subjects were classified into four groups. Group 4 members were associated with relatively higher oxygen consumption rates.