Xue-Wei Tian

New algorithm for the depression diagnosis using HRV: A neuro-fuzzy approach

Recent research indicates a significant relationship between the severity of depression and heart rate variability (HRV). This paper presents a neuro-fuzzy approach-based classification algorithm, which distinguishes patients with depression from controls by a neuro-fuzzy network with a weighted fuzzy membership function (NEWFM) using the two time domain and four frequency domain features of HRV. The HRV data were collected from 10 patients with depression and an equal number of healthy controls. Wearing a wireless Holter monitor, each subject underwent a 13-minute multimodal affective contents stimulus, which can induce a variety of emotions. HRV activity was transformed and recorded from periods of 13-minute ECG signals. With a reliable accuracy rate of 95%, the six HRV features were extracted and used as NEWFM input features for depression classification. The standard deviation of the RR intervals (SDNN) and very low frequency (VLF) of HRV were evaluated as good features-from six features-by a non-overlap area distribution measurement method. The two features reflected conspicuous differences between the depression diagnosed and the healthy subjects, which indicates a significant association between depression and the autonomic nervous system. The proposed algorithm will be implemented as a depression monitoring system in a Smartphone application.

Real-time algorithm for a mobile cardiac monitoring system to detect life-threatening arrhythmias

This paper presents a real-time algorithm for a mobile cardiac monitoring system to detect life-threatening arrhythmias. This detection algorithm focuses on two life-threatening arrhythmias ventricular tachycardia and fibrillation (VT/VF), which are detected through the application of pre-detection processing and main detection processing. In pre-detection processing, applies a statistical method to detect VT/VF. In contrast, a neural fuzzy network is applied to detect VT/VF in main detection processing. The neural fuzzy networks input features are obtained by wavelet transform and several effective extraction methods. This real-time detection algorithm outperform Amanns algorithm, with 92% accuracy and 93% sensitivity. It has been implemented as a cardiac monitoring system in a mobile phone. This system meets heart patients requirements of early detection and out-of-hospital rehabilitation.

A 2-D Visual Model for Sasang Constitution Classification Based on a Fuzzy Neural Network

The human constitution can be classified into four possible constitutions according to an individual’s temperament and nature: Tae-Yang (太陽), So-Yang (少陽), Tae-Eum (太陰), and So-Eum (少陰). This classification is known as the Sasang constitution. In this study, we classified the four types of Sasang constitutions by measuring twelve sets of meridian energy signals with a Ryodoraku device (良導絡). We then developed a Sasang constitution classification method based on a fuzzy neural network (FNN) and a two-dimensional (2-D) visual model. We obtained meridian energy signals from 35 subjects for the So-Yang, Tae-Eum, and So-Eum constitutions. A FNN was used to obtain defuzzification values for the 2-D visual model, which was then applied to the classification of these three Sasang constitutions. Finally, we achieved a Sasang constitution recognition rate of 89.4 %.

Learning Distribution Graphs Using a Neuro-Fuzzy Network for Naive Bayesian Classifier

Abstract Naive Bayesian classifiers are a powerful and well-known type of classifiers that can be easily induced from a dataset of sample cases. However, the strong conditional independence assumptions can sometimes lead to weak classification performance. Normally, naive Bayesian classifiers use Gaussian distributions to handle continuous attributes and to represent the likelihood of the features conditioned on the classes. The probability density of attributes, however, is not always well fitted by a Gaussian distribution. Another eminent type of classifier is the neuro-fuzzy classifier, which can learn fuzzy rules and fuzzy sets using supervised learning. Since there are specific structural similarities between a neuro-fuzzy classifier and a naive Bayesian classifier, the purpose of this study is to apply learning distribution graphs constructed by a neuro-fuzzy network to naive Bayesian classifiers. We compare the Gaussian distribution graphs with the fuzzy distribution graphs for the naive Bayesian classifier. We applied these two types of distribution graphs to classify leukemia and colon DNA microarray data sets. The results demonstrate that a naive Bayesian classifier with fuzzy distribution graphs is more reliable than that with Gaussian distribution graphs.

Emotional-speech recognition using the neuro-fuzzy network

Emotion recognition based on a speech signal is one of the intensively studied research topics in the domains of human-computer interaction and affective computing. The presented paper is concerned with emotional-speech recognition based on the neuro-fuzzy network with a weighted fuzzy membership function (NEWFM). NEWFM has a feature selection method and makes fuzzy classifiers. In this paper, NEWFM was utilized for classifying four kinds of emotional-speech signals. This NEWFM classification method achieves as high as 86% overall classification accuracy. Significantly, the NEWFM classifier efficiently detects sadness, with a 97.5% recognition rate.

A Prototype Selection Algorithm Using Fuzzy k-Important Nearest Neighbor Method

The k-Nearest Neighbor (KNN) algorithm is widely used as a simple and effective classification algorithm. While its main advantage is its simplicity, its main shortcoming is its computational complexity for large training sets. A Prototype Selection (PS) method is used to optimize the efficiency of the algorithm so that the disadvantages can be overcome. This paper presents a new PS algorithm, namely Fuzzy k-Important Nearest Neighbor (FKINN) algorithm. In this algorithm, an important nearest neighbor selection rule is introduced. When classifying a data set with the FKINN algorithm, the most repeated selection sample is defined as an important nearest neighbor. To verify the performance of the algorithm, five UCI benchmarking databases are considered. Experiments show that the algorithm effectively deletes redundant or irrelevant prototypes while maintaining the same level of classification accuracy as that of the KNN algorithm.

Depression and Fatigue Analysis Using a Mental-Physical Model

Recent research has indicated a significant association between depression and fatigue. To analyze depression and fatigue, an experiment was conducted that provided the subjects with affective content to induce a variety of emotions and heart rate variability (HRV). This paper presents a mental–physical model that describes the relationship between depression and fatigue by using a neuro-fuzzy network with a weighted fuzzy membership function using two time-domain and four frequency-domain features of HRV. HRV data were collected from 24 patients. At the end of the experiment, we determined the relationship between depression and fatigue with the mental–physical model, and our analysis results had an accuracy of 95.8 %.

Comparison of HRV Time and Frequency Domain Features for Myocardial Ischemia Detection

Heart Rate Variability (HRV) analysis is a convenient tool to assess Myocardial Ischemia (MI). The analysis methods of HRV can be divided into time domain and frequency domain analysis. This paper uses wavelet transform as frequency domain analysis in contrast to time domain analysis in short term HRV analysis. ST-T and normal episodes are collected from the European ST-T database and the MIT-BIH Normal Sinus Rhythm database, respectively. An episode can be divided into several segments, each of which is formed by 32 successive RR intervals. Eighteen HRV features are extracted from each segment by the time and frequency domain analysis. To diagnose MI, the Neural Network with Weighted Fuzzy Membership functions (NEWFM) is used with the extracted 18 features. The results show that the average accuracy from time and frequency domain features is 75.29% and 80.93%, respectively.