The Use of Time-Frequency Moments as Inputs of LSTM Network for ECG Signal Classification

Abstract

This paper refers to the method of using the deep neural long-short-term memory (LSTM) network for the problem of electrocardiogram (ECG) signal classification. ECG signals contain a lot of subtle information analyzed by doctors to determine the type of heart dysfunction. Due to the large number of signal features that are difficult to identify, raw ECG data is usually not suitable for use in machine learning. The article presents how to transform individual ECG time series into spectral images for which two characteristics are determined, which are instantaneous frequency and spectral entropy. Feature extraction consists of converting the ECG signal into a series of spectral images using short-term Fourier transformation. Then the images were converted using Fourier transform again to two signals, which includes instantaneous frequency and spectral entropy. The data set transformed in this way was used to train the LSTM network. During the experiments, the LSTM networks were trained for both raw and spectrally transformed data. Then, the LSTM networks trained in this way were compared with each other. The obtained results prove that the transformation of input signals into images can be an effective method of improving the quality of classifiers based on deep learning.

Recognition of Emotion in Indian Classical Dance Using EMG Signal

“…This research selected 373 hidden nodes for maximum classification accuracy and the 'Adam' algorithm for training. Adam is an adaptive learning rate optimization algorithm [30]- [32].…”

“…10 (a) and (b) show four channels IEMG features for all nine emotions of subjects 2 and 3. Slices represent (0-18) for Adbhut (amazed), (19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29)(30)(31)(32)(33)(34)(35)(36)(37) for Bhayanaka (fearful), (38-56) for Hasya (humorous), (57-75) for Karuna (tragic), (76-94) for Raudra (fierce), (95-113) for Shant (peaceful), (114)(115)(116)(117)(118)(119)(120)(121)(122)(123)(124)(125)(126)(127)(128)(129)(130)(131)(132) for Shringar (loving smile), (133-151) for Veer (heroic), and (152-170) for Bibhatsa (disgusted). Each subject has its IEMG feature value.…”

Classification of COVID-19 electrocardiograms by using hexaxial feature mapping and deep learning

“…Additionally, many deep learning approaches were proposed for automatic cardiac arrhythmia detection. Besides using 1D ECG signals [ 38 , 39 ] to train the deep network, in many studies were used a 2D representation of 1D ECG signals like ECG time-amplitude images [ 40 – 43 ], time-frequency representations by using Short-Time Fourier Transform (STFT) [ 44 , 45 ] and Continuous Wavelet Transform (CWT) [ 46 ], higher-order spectral representations [ 47 ], and dual beat coupling matrices [ 48 ] in order to train CNN architecture. Considering the wide usage of paper-based ECG reports [ 49 ], there is a lack in the automatic detection of cardiac problems which require special attention.…”

Classification of COVID-19 electrocardiograms by using hexaxial feature mapping and deep learning

“…Besides using 1D ECG signals [38,39] to train the deep network, in many studies were used a 2D representation of 1D ECG signals like ECG time-amplitude images [40,41], time-frequency representations by using Short-Time Fourier Transform (STFT) [42,43] and Continuous Wavelet Transform (CWT) [44], higherorder spectral representations [45], and dual beat coupling matrices [46] in order to train CNN architecture.…”

“…3)43 We assessed the four GLCM features from a statisti-44 cal perspective in order to select the most informative45 and distinctive feature to represent the binary ECG 46 images. We performed the one-way ANOVA test on 47 GLCM features obtained from the binary ECG images 48.…”