基于混合时频域特征的卷积神经网络心律失常分类方法的研究

心律失常是常见的心血管疾病之一，目前很多方法通过计算机辅助系统对心电图进行分析以识别心律失常，但由于大多数心律失常数据样本较少，计算机辅助系统识别心律失常效果不佳.本文提出了一种基于混合时频域分析特征提取的卷积神经网络方法，该方法提取心电图的RR间期时域特征、希尔伯特-黄变换提取的频域特征和连续小波变换提取的时频域联合特征，经过特征融合后输入卷积神经网络训练分类模型，并采用Focal Loss作为网路的损失函数，实现对心律失常的分类.本文使用MIT-BIH(Massachusetts Institute of Technology-Boston’s Beth Israel Hospital)心律失常数据库验证本文提出方法对4类心电数据分类的结果，实验结果表明，与现有的分类算法相比，本文所提出的混合时频域特征方法能有效提升心律失常分类的准确性.     

基于滤波重构和卷积神经网络的心电信号分类

计算机自动分类心电信号能够减轻医生工作压力并大幅提高诊断速度和准确率。文中针对传统算法中特征提取过程复杂及抗干扰能力弱的问题,提出了一种结合滤波重构和卷积神经网络的心电信号分类算法。该算法首先通过传统信号滤波和心拍序列重构去除原始心电信号中的噪声干扰,然后构建卷积神经网络来自动学习心电信号特征并完成分类。在PhysioNet/CinC Challenge 2017数据集上的分类实验结果表明,该方法的平均F₁(查准率、召回率的调和平均)达到了0.8471,优于人工特征提取和常规卷积网络方法,且具有很强的抗干扰能力。     

基于形态特征提取的急性下壁心肌梗死BiLSTM网络辅助诊断算法

急性下壁心肌梗死是一种病发急、进展快、致死率高的心脏疾病，该文提出一种新颖的基于形态特征提取的BiLSTM神经网络分类的急性下壁心肌梗死辅助诊断算法，可大幅度提高医生对急性下壁心肌梗死疾病的诊断效率并有助于及时确诊。算法包括:对胸痛中心数据库心拍信号进行降噪及心拍分割；根据临床心内科医学诊断指南提取了12导联波形距离特征和分导联波形幅值特征；依据提取的特征搭建LSTM与BiLSTM神经网络进行心拍的分类识别；使用PTB公开数据库和胸痛中心数据库多临床中心进行交叉验证。实验结果表明，加入胸痛中心真实临床数据后，基于形态特征提取BiLSTM神经网络的急性下壁心肌梗死辅助诊断算法准确率达到99.72%，精度达到99.53%，灵敏度达到100.00%，同时F1-Score达到99.76。该算法比其他现有算法准确率提高至少1%，该项研究具有非常重要的临床应用价值。     

基于形态特征提取的急性下壁心肌梗死BiLSTM网络辅助诊断算法

急性下壁心肌梗死是一种病发急、进展快、致死率高的心脏疾病,该文提出一种新颖的基于形态特征提取的BiLSTM神经网络分类的急性下壁心肌梗死辅助诊断算法,可大幅度提高医生对急性下壁心肌梗死疾病的诊断效率并有助于及时确诊.算法包括:对胸痛中心数据库心拍信号进行降噪及心拍分割;根据临床心内科医学诊断指南提取了12导联波形距离特征和分导联波形幅值特征;依据提取的特征搭建LSTM与BiLSTM神经网络进行心拍的分类识别;使用PTB公开数据库和胸痛中心数据库多临床中心进行交叉验证.实验结果表明,加入胸痛中心真实临床数据后,基于形态特征提取BiLSTM神经网络的急性下壁心肌梗死辅助诊断算法准确率达到99.72％,精度达到99.53％,灵敏度达到100.00％,同时F1-Score达到99.76.该算法比其他现有算法准确率提高至少1％,该项研究具有非常重要的临床应用价值.     

加权联合降维的深度特征提取与分类识别算法

为了降低卷积神经网络计算的复杂度,改善特征提取过程中的过拟合现象,解决经典网络模型不能有效处理大尺寸图片的问题,采用了加权联合降维的特征融合与分类识别算法,根据两特征的识别贡献率对主成分分析法（PCA）降维处理和随机投影（RP）处理结果进行加权融合,然后将结果提供给卷积神经网络进行处理,提取图像分类的高层特征,使用欧氏距离分类器对识别对象进行分类,并进行了理论分析和实验验证。结果表明,经过加权联合降维对数据进行预处理,PCA矩阵与RP降维矩阵之比重达到6:4,识别率高达96%以上。该算法有效提高了准确率,使大尺寸图片在深度学习网络中有良好的识别效果,改善了网络的适应性。     

基于信噪比分类网络的调制信号分类识别算法

针对传统降噪算法损伤高信噪比(SNR)信号而造成信号识别准确率下降的问题,该文提出基于卷积神经网络的信噪比分类算法,该算法利用卷积神经网络对信号进行特征提取,用固定K均值(FK-means)算法对提取的特征进行聚类处理,准确分类高低信噪比信号。低信噪比信号采用改进的中值滤波算法降噪,改进的中值滤波算法在传统中值滤波的基础上增加了前后采样窗口的关联性机制,来改善传统中值滤波算法处理连续噪声效果不佳的问题。为充分提取信号的空间特征和时间特征,该文提出卷积神经网络和长短时记忆网络并联的卷积长短时(P-CL)网络,利用卷积神经网络和长短时记忆网络分别提取信号的空间特征与时间特征,并进行特征融合与分类。实验表明,该文提出的调制信号分类模型识别准确率为91%,相比于卷积长短时(CNN-LSTM)网络提高了6%。     

分离通道联合卷积神经网络的自动调制识别

针对通信信号的自动调制识别需要大量特征提取的问题,提出了一种分离通道卷积神经网络自动调制识别算法。该算法通过结合深度学习中卷积神经网络(CNN),分别提取时域信号的多通道和分离通道调制特征,再利用融合特征实现不同信号的分类。仿真结果表明,相比基于CNN的算法,所提算法在高信噪比下针对两个数据集的识别率分别提升7%和18%;此外,相比于基于特征提取的传统识别算法,其高阶调制识别性能平均提升3 dB。     

基于多维度特征遥感图像分类方法的研究北大核心CSCD

为了解决传统高光谱图像分类方法精度低、计算成本高及未能充分利用空-谱信息的问题,本文提出一种基于多维度并行卷积神经网络(multidimensional parallel convolutional neural network,3D-2D-1D PCNN)的高光谱图像分类方法。首先,该算法利用不同维度卷积神经网络(convolutional neural network,CNN)提取高光谱图像信息中的空-谱特征、空间特征及光谱特征;之后,采用相同并行卷积层将组合后的空-谱特征、空间特征及光谱特征进行特征融合;最后,通过线性分类器对高光谱图像信息进行精准分类。本文所提方法不仅可以提取高光谱图像中更深层次的空间特征和光谱特征信息,同时能够将光谱图像不同维度的特征进行融合,减小计算成本。在Indian Pines、Pavia Center和Pavia University数据集上对本文算法和4种传统算法进行对比实验,结果表明,本文算法均得到最优结果,分类精度分别达到了99.210%、99.755%和99.770%。     

核卷积神经网络研究与应用

利用核函数非线性映射的优势,结合卷积神经网络算法,提出一种基于核卷积神经网络（Kernel-Convolutional Neural Network , Kernel-CNN）的新的网络学习模型。该方法首先对数据预处理,其次利用核卷积神经网络对数据进行特征提取,最后,构建softmax分类器对数据进行分类。本网络将非线性映射引入卷积过程构成核卷积过程,通过核卷积过程进一步增强模型的特征提取能力,在MNIST手写数字库以及美国麻省理工学院提供的MIT-BIH心律失常数据库上实验验证,本文模型正确率分别为98.5%、97%,均较好于卷积神经网络和支持向量机,且本文模型具有较小的LOSS值。      

基于TensorFlow卷积神经网络的图像识别

TensorFlow是Google公司发布的开源人工智能深度学习框架,卷积神经网络是进行图像识别的一种有效方法。本文在研究Tensorflow深度学习框架以及卷积神经网络的基础上,利用keras官方下载的cifar数据集,采用LeNet-5算法对数据进行了处理、建模、训练、并对模型进行了评估以及保存,利用测试集完成测试后,不同图像识别的准确率有所不同,青蛙识别的准确率最高,为79%,汽车的识别准确率为78%,猫和狗的识别准确率最低,分别为41%和53%,所有图像识别的平均准确率为65%。     

Detection of premature ventricular contraction arrhythmias in electrocardiogram signals with kernel methods

In this paper, we propose to investigate the capabilities of two kernel methods for the detection and classification of premature ventricular contractions (PVC) arrhythmias in Electrocardiogram (ECG signals). These kernel methods are the support vector machine and Gaussian process (GP). We propose to study these two classifiers with various feature representations of ECG signals, such as morphology, discrete wavelet transform, higher-order statistics, and S transform. The experimental results obtained on 48 records (i.e., 109,887 beats) of the MIT-BIH Arrhythmia database showed that for all feature representation adopted in this work, the GP detector trained only with 600 beats from PVC and Non-PVC classes can provide an overall accuracy and a sensitivity above 90 % on 20 records (i.e., 49,774 beats) and 28 records (i.e., 60,113 beats) seen and unseen, respectively, during the training phase.     

An association rule mining-based methodology for automated detection of ischemic ECG beats

Currently, an automated methodology based on association rules is presented for the detection of ischemic beats in long duration electrocardiographic (ECG) recordings. The proposed approach consists of three stages. 1) Preprocessing: Noise is removed and all the necessary ECG features are extracted. 2) Discretization: The continuous valued features are transformed to categorical. 3) Classification: An association rule extraction algorithm is utilized and a rule-based classification model is created. According to the proposed methodology, electrocardiogram (ECG) features extracted from the ST segment and the T-wave, as well as the patient's age, were used as inputs. The output was the classification of the beat as ischemic or not. Various algorithms were tested both for discretization and for classification using association rules. To evaluate the methodology, a cardiac beat dataset was constructed using several recordings of the European Society of Cardiology ST-T database. The obtained sensitivity (Se) and specificity (Sp) was 87% and 93%, respectively. The proposed methodology combines high accuracy with the ability to provide interpretation for the decisions made, since it is based on a set of association rules.     

A low-complexity data-adaptive approach for premature ventricular contraction recognition

Premature ventricular contraction (PVC) may lead to life-threatening cardiac conditions. Real-time automated PVC recognition approaches provide clinicians the useful tools for timely diagnosis if dangerous conditions surface in their patients. Based on the morphological differences of the PVC beats in the ventricular depolarization phase (QRS complex) and repolarization phase (mainly T-wave), two beat-to-beat template-matching procedures were implemented to identify them. Both templates were obtained by a probability-based approach and hence were fully data-adaptive. A PVC recognizer was then established by analyzing the correlation coefficients from the two template-matching procedures. Our approach was trained on 22 ECG recordings from the MIT-BIH arrhythmia database (MIT-BIH-AR) and then tested on another 22 nonoverlapping recordings from the same database. The PVC recognition accuracy was 98.2 %, with the sensitivity and positive predictivity of 93.1 and 81.4 %, respectively. To evaluate its robustness against noise, our approach was applied again to the above testing set, but this time, the ECGs were not preprocessed. A comparable performance was still obtained. A good generalization capability was also confirmed by validating our approach on an independent St. Petersburg Institute of Cardiological Technics database. In addition, our performance was comparable with these published complex approaches. In conclusion, we have developed a low-complexity data-adaptive PVC recognition approach with good robustness against noise and generalization capability. Its performance is comparable to other state-of-the-art methods, demonstrating a good potential in real-time application.     

Effect of Multiscale PCA De-noising in ECG Beat Classification for Diagnosis of Cardiovascular Diseases

Current trends in clinical applications demand automation in electrocardiogram (ECG) signal processing and heart beat classification. This paper examines the design of an effective recognition method to diagnose heart diseases. The proposed method consists of three main modules: de-noising module, feature extraction module, and classifier module. In the de-noising module, multiscale principal component analysis (MSPCA) is used for noise reduction of the ECG signals. In the feature extraction module, autoregressive (AR) modeling is used for extracting features. In the classifier module, different classifiers are examined such as simple logistic, k-nearest neighbor, multilayer perceptron, radial basis function networks, and support vector machines. Different experiments are carried out using the MIT-BIH arrhythmia database to classify different ECG heart beats and the performance of the proposed method is evaluated in terms of several standard metrics. The experimental results show that the proposed method is able to reduce noise from the noisy ECG signals more accurately in comparison to previous methods. The numerical results indicated that the proposed algorithm achieved 99.93 % of the classification accuracy using MSPCA de-noising and AR modeling.     

ECG beat recognition using fuzzy hybrid neural network

This paper presents the application of the fuzzy neural network for electrocardiographic (ECG) beat recognition and classification. The new classification algorithm of the ECG beats, applying the fuzzy hybrid neural network and the features drawn from the higher order statistics has been proposed in the paper. The cumulants of the second, third, and fourth orders have been used for the feature selection. The hybrid fuzzy neural network applied in the solution consists of the fuzzy self-organizing subnetwork connected in cascade with the multilayer perceptron, working as the final classifier. The c-means and Gustafson-Kessel algorithms for the self-organization of the neural network have been applied. The results of experiments of recognition of different types of beats on the basis of the ECG waveforms have confirmed good efficiency of the proposed solution. The investigations show that the method may find practical application in the recognition and classification of different type heart beats.     

Analysis of first-derivative based QRS detection algorithms.

Accurate QRS detection is an important first step for the analysis of heart rate variability. Algorithms based on the differentiated ECG are computationally efficient and hence ideal for real-time analysis of large datasets. Here, we analyze traditional first-derivative based squaring function (Hamilton-Tompkins) and Hilbert transform-based methods for QRS detection and their modifications with improved detection thresholds. On a standard ECG dataset, the Hamilton-Tompkins algorithm had the highest detection accuracy (99.68% sensitivity, 99.63% positive predictivity) but also the largest time error. The modified Hamilton-Tompkins algorithm as well as the Hilbert transform-based algorithms had comparable, though slightly lower, accuracy; yet these automated algorithms present an advantage for real-time applications by avoiding human intervention in threshold determination. The high accuracy of the Hilbert transform-based method compared to detection with the second derivative of the ECG is ascribable to its inherently uniform magnitude spectrum. For all algorithms, detection errors occurred mainly in beats with decreased signal slope, such as wide arrhythmic beats or attenuated beats. For best performance, a combination of the squaring function and Hilbert transform-based algorithms can be applied such that differences in detection will point to abnormalities in the signal that can be further analyzed.     

QRS feature extraction using linear prediction

This communication proposes a method called linear prediction (a high performant technique in digital speech processing) for analyzing digital ECG signals. There are several significant properties indicating that ECG signals have an important feature in the residual error signal obtained after processing by Durbin's linear prediction algorithm. This communication also indicates that the prediction order need not be more than two for fast arrhythmia detection. The ECG signal classification puts an emphasis on the residual error signal. For each ECG's QRS complex, the feature for recognition is obtained from a nonlinear transformation which transforms every residual error signal to a set of three states pulse-code train relative to the original ECG signal. The pulse-code train has the advantage of easy implementation in digital hardware circuits to achieve automated ECG diagnosis. The algorithm performs very well in feature extraction in arrhythmia detection. Using this method, our studies indicate that the PVC (premature ventricular contraction) detection has at least a 92 percent sensitivity for MIT/BIH arrhythmia database.     

基于集合经验模态分解和信号结构分析的心电信号R波识别算法

R波作为确定心电信号各波段的重要参考,是心电自动分析的前提。针对大多数R波识别算法的预处理过程影响识别准确度和耗时问题,该文提出一种基于集合经验模态分解(EEMD)和信号结构分析的算法对带噪心电信号(ECG)的R波直接进行识别。首先通过EEMD将带噪声的心电信号分解成一系列本征模态分量,然后对分解后的各模态分量作独立成分分析以提取出R波特征最明显的成分,对该成分进行结构分析,从而实现对R波的准确定位。仿真结果表明,该文算法对带噪声心电信号的R波识别具有更优性能,对异常心电信号的R波识别也具有明显效果。     

基于三轴传感器的老年人日常活动识别

本文针对老年人日常活动类型及特点提出了一种基于三轴加速度传感器和HMM（Hidden Markov Model）的活动识别方法.本文首先提取了针对老年人相异、相似活动的标准差、能量、相关系数、RAF（RAtio Forward）、RVF（Ratio Vertical Forward）等特征值.然后定义老年人的HMM活动识别模型.最后在经过Baum-Welch算法对HMM进行参数训练后使用Viterbi算法来进行老年人活动识别.实验结果表明,本文方法适用于老年人的日常活动的识别,平均识别精度达到了93.3%,尤其是对于相似步态活动的识别准确率达到了93.7%.     

基于改进深度生成对抗网络的心电信号重构算法

心冲击图(BCG)信号中含有睡眠时期的心跳等生理参数,采用非接触式测量,但易受干扰影响应用受限;心电图(ECG)信号应用很广,但采用接触式测量,操作不便.为了实现非接触式测量并监测心电信号,该文将无参数尺度空间法(PSA)引入并与经验小波变换(EWT)算法结合,从BCG信号中分解得到心跳分量,结果表明所提分解方法能有效...