In recent years, there has been an explosion of interest in mining time-series databases. As with most computer science problems, representation of the data is the key to efficient and effective solutions. One of the most commonly used representations is piecewise linear approximation. This representation has been used by various researchers to support clustering, classification, indexing and association rule mining of time-series data. A variety of algorithms have been proposed to obtain this representation, with several algorithms having been independently rediscovered several times. In this paper, we undertake the first extensive review and empirical comparison of all proposed techniques. We show that all these algorithms have fatal flaws from a data-mining perspective. We introduce a novel algorithm that we empirically show to be superior to all others in the literature.

/pdf/an-online-algorithm-for-segmenting-time-series-yukam6bgmy.pdf

An online algorithm for segmenting time series

Electrocardiogram (ECG) compression techniques are compared, and a unified view of these techniques is established. ECG data compression schemes are presented in two major groups: direct data compression and transformation methods. The direct data compression techniques are ECG differential pulse code modulation (DPCM) and entropy coding, AZTEC, Turning-point, CORTES, Fan and SAPA algorithms, peak-picking, and cycle-to-cycle compression methods. The transformation methods include Fourier, Walsh, and Karhunen-Loeve transforms. The theoretical bases behind the direct ECG data compression schemes are presented and classified into three categories: tolerance-comparison compression, DPCM, and entropy coding methods. A framework for evaluation and comparison of ECG compression schemes is presented. >

ECG data compression techniques-a unified approach

In recent years, there has been an explosion of interest in mining time series databases. As with most computer science problems, representation of the data is the key to efficient and effective solutions. One of the most commonly used representations is piecewise linear approximation. This representation has been used by various researchers to support clustering, classification, indexing and association rule mining of time series data. A variety of algorithms have been proposed to obtain this representation, with several algorithms having been independently rediscovered several times. In this paper, we undertake the first extensive review and empirical comparison of all proposed techniques. We show that all these algorithms have fatal flaws from a data mining perspective. We introduce a novel algorithm that we empirically show to be superior to all others in the literature.

/pdf/segmenting-time-series-a-survey-and-novel-approach-3yq0qskz81.pdf

Segmenting Time Series: A Survey and Novel Approach

In this paper, a new distortion measure for electrocardiogram (ECG) signal compression, called weighted diagnostic distortion (WDD) is introduced. The WDD measure is designed for comparing the distortion between original ECG signal and reconstructed ECG signal (after compression). The WDD is based on PQRST complex diagnostic features (such as P wave duration, QT interval, T shape, ST elevation) of the original ECG signal and the reconstructed one. Unlike other conventional distortion measures [e.g. percentage root mean square (rms) difference, or PRD], the WDD contains direct diagnostic information and thus is more meaningful and useful. Four compression algorithms were implemented (AZTEC, SAPA2, LTP, ASEC) in order to evaluate the WDD. A mean opinion score (MOS) test was applied to test the quality of the reconstructed signals and to compare the quality measure (MOS/sub error/) with the proposed WDD measure and the popular PRD measure. The evaluators in the WIGS test were three independent expert cardiologists, who studied the reconstructed ECG signals in a blind and a semiblind tests. The correlation between the proposed WDD measure and the MOS test measure (MOS/sub error/) was found superior to the correlation between the popular PRD measure and the MOS/sub error/.

The weighted diagnostic distortion (WDD) measure for ECG signal compression

A Survey of Recent Methods for Efficient Retrieval of Similar Time Sequences (H M Lie) Indexing of Compressed Time Series (E Fink & K Pratt) Boosting Interval-Based Literal: Variable Length and Early Classification (J J Rodriguez Diez) Segmenting Time Series: A Survey and Novel Approach (E Keogh et al) Indexing Similar Time Series under Conditions of Noise (M Vlachos et al) Classification of Events in Time Series of Graphs (H Bunke & M Kraetzl) Median Strings - A Review (X Jiang et al) Change Detection in Classification Models of Data Mining (G Zeira et al).

Data Mining in Time Series Databases

Three fast and efficient "scan-along" algorithms for compressing digitized electrocardiographic data are described. These algorithms are "scan-along" in the sense that they produce the compressed data in real time as the electrocardiogram is generated. The algorithms are based on the minimum perimeter polygonal approximation for digitized curves. The approximation restricts the maximum error to be no greater than a specified value. Our algorithms achieve a compression ratio of ten on a database of 8000 5-beat abnormal electrocardiograms sampled at 250 Hz and a compression ratio of eleven on a database of 600 3-beat normal electrocardiograms (different from the preceding database) sampled at 500 Hz.

Scan-Along Polygonal Approximation for Data Compression of Electrocardiograms

In this research, we proposed that high frequency component of HFECG was applicable biometric feature for new identification system We developed identification method by using neural network (NN), and aimed at the improvement of the classification rate Preprocessing prior to NN is performed by justification on time axis and normalization on amplitude As a result, an average of 99% classification rate was obtained from 9 subjects We also made an attempt to identify in shorter time by shifting of the HFECG by a few samples to NN

Individual identification with high frequency ECG : Preprocessing and classification by neural network

Although there is an invasive method, the procedure to quantitate the muscle fatigue today is not fully established This study analyzes the muscle fatigue by measuring cross electric property of the muscle with working load noninvasively We measured the electrical impedance of the muscle to obtain the relationship between the electrical property and the fatigue during exercise Several Ag/AgCl electrodes are placed the biceps brachii muscle crossing the arm The baseline drift due to the sweat glands on the skin was effectively eliminated by subtracting one of the exercised arm from one of the resting arm, and the interelectrode impedance was measured The fatigue of the arm was loaded by flexing exercise with a 2 kg iron array The rate of flexing was once per second The impedance was acquired every 30 seconds The flexing was constant but ceased while acquiring the impedance value for 5 seconds As a result impedance increased at the beginning, and then decreased after 90 seconds, or started decreasing after the beginning of the exercise These phenomena were well reproducible The initial increase of the impedance was concluded as with the fatigue which may triggered the increase of the lactic acid in the muscle The decrease of the impedance in the latter part was probably due to the influence of increased blood flow, which was simultaneously measured in this study

Dynamic Electrical Characteristics of the Muscle with the Fatigue during Exercise

This paper compares the performance of three methods to reconstruct underlying heart rate from observed ECG pulse trains. Artificial integrated pulse frequency modulated pulse trains with 1/f fluctuations, which simulate ECG pulse trains, were generated and used for the evaluation of reconstruction methods. Instantaneous heart rate, the heart rate reconstructed by the Berger's method and the derivative of cubic spline interpolation (DCSI) method were compared. The performances of those methods were examined by comparing the spectral pattern of reconstructed heart rate signals. The DCSI method was shown to give the best performance preserving 1/f spectral characteristics over the wide frequency range.

http://ieeexplore.ieee.org/iel5/9009/28608/01279591.pdf

Reconstruction of underlying heart rate from integrated pulse frequency modulated heart pulses with 1/f fluctuations

The dc component of biological sources is difficult to measure with body surface electrodes. This is due to the existence of an electrical polarization in the interface between the electrode and the electrolyte. A system to acquire a dc component of the biological origin is proposed by means of differentiating the polarization potential from the biological one. The acquisition system is comprised of an instrumentation amplifier, an integrated semiconductor analog switch, and ordinary silver/silver-chloride body surface electrodes. Biopotentials, electrooculograms from the eyes, were noninvasively recorded by the system with the analog switch which was inserted between the electrodes to make the two inputs electrically open following them being shorted. Asymptotes of a triple exponential function were subsequently extracted by regression model analysis on the output signals that were acquired immediately after the switch opened. One of the distinctive asymptotes of the function showed that the magnitude which corresponded to eye movements stayed stable as long as the eyes were fixed. Analysis of the time constants involved in the acquisition system elucidated that the asymptote was the biological origin.

Masa Ishijima

Papers

Scan-Along Polygonal Approximation for Data Compression of Electrocardiograms

Individual identification with high frequency ECG : Preprocessing and classification by neural network

Dynamic Electrical Characteristics of the Muscle with the Fatigue during Exercise

Reconstruction of underlying heart rate from integrated pulse frequency modulated heart pulses with 1/f fluctuations

Acquisition of DC Component of Biopotentials With Body Surface Electrodes