Deep Belief Network Classifier Research Articles

SAR-BSO meta-heuristic hybridization for feature selection and classification using DBNover stream data

Advancements in cloud technologies have increased the infrastructural needs of data centers due to storage needs and processing of extensive dimensional data. Many service providers envisage anomaly detection criteria to guarantee availability to avoid breakdowns and complexities caused due to large-scale operations. The streaming log data generated is associated with multi-dimensional complexity and thus poses a considerable challenge to detect the anomalies or unusual occurrences in the data. In this research, a hybrid model is proposed that is motivated by deep belief criteria and meta-heuristics. Using Search-and-Rescue—BrainStorm Optimization (SAR-BSO), a hybrid feature selection (FS) and deep belief network classifier is used to localize and detect anomalies for streaming data logs. The significant contribution of the research lies in FS, which is carried out using SAR-BSO which increases the detection power of the model as it selects the most significant variables by minimizing redundant features. The evaluation of accuracy is efficiently improved when compared with the predictable methods, such as Extract Local Outlier Factor (ELOF), Track-plus, Hybrid Distributed Batch Stream (HDBS), IForestASD, DBN, BSO-based Feature Selection with DBN, Genetic Algorithm-Deep Belief Network (GA-DBN), Mutual Information-Deep Belief Network (MI-DBN), information entropy-Deep Belief Network(I + DBN), Flat Field-Deep Belief Network (FF + DBN), African Vulture Optimization Algorithm-Deep Belief Network(AVOA + DBN), Gorilla Troop Optimizer-Deep Belief Network(GTO-DBN), and SARO-based Feature Selection with DBN. Further, the accurate detection of the anomalies in the data stream is established by the Deep Belief Neural Network (DBN) classifier. The model’s efficacy is determined using Apache, Hadoop, HDFS, Spark, and Linux datasets and evaluated against existing similar models. The model efficiency is provided using multiple evaluation metrics and is found effective. From the experimentation, the accuracy of the proposed model is found to be 93.3, 95.4, 93.6, 94.2, and 93.5% respectively for the dataset such as Apache, Hadoop, HDFS, spark, and Linux. This enhancement in accuracy is due to the selection of optimal features by the proposed SAR-BSO algorithm.

Comparative analysis of image classification with retrieval system

Currently, the term 'content-based image retrieval' seems to be a highly attentive system for handling the broad image datasets since the data storage mechanisms and image acquisition are becoming the most empowered logic in image processing. The previous CBIR system has been proposed under nonlinear similarity matching measure in a logarithmic scale and informative pattern descriptor has quantified the range of similarity content. This article implements a novel CBIR system that emphasises the classification concept using a deep belief network (DBN) classifier. In this concept, apart from the image retrieval, the used classifier classifies the respective classes of retrieved images. Finally, the proposed local vector pattern (PLVP) with DBN classifier (PLVP-DBN) compares its performance over other conventional retrieval concepts: PLVP-with log similarity, PLVP-without log similarity, and also with neural network (NN) classifier.

Enhanced Optimized Classification Model of Chronic Kidney Disease

Chronic kidney disease (CKD) is one of the leading causes of death across the globe, affecting about 10% of the world's adult population. Kidney disease affects the proper function of the kidneys. As the number of people with chronic kidney disease (CKD) rises, it is becoming increasingly important to have accurate methods for detecting CKD at an early stage. Developing a mechanism for detecting chronic kidney disease is the study's main contribution to knowledge. In this study, preventive interventions for CKD can be explored using machine learning techniques (ML). The Optimized deep belief network (DBN) based on Grasshopper's Optimization Algorithm (GOA) classifier with prior Density-based Feature Selection (DFS) algorithm for chronic kidney disease is described in this study, which is called "DFS-ODBN." Prior to the DBN classifier, whose parameters are optimized using GOA, the proposed method eliminates redundant or irrelevant dimensions using DFS. The proposed DFS-ODBN framework consists of three phases, preprocessing, feature selection, and classification phases. Using CKD datasets, the suggested approach is also tested, and the performance is evaluated using several assessment metrics. Optimized-DBN achieves its maximum performance in terms of sensitivity, accuracy, and specificity, the proposed DFS-ODBN demonstrated accuracy of 99.75 percent using fewer features comparing with other techniques.

A Method Combining Multi-Feature Fusion and Optimized Deep Belief Network for EMG-Based Human Gait Classification

In this paper, a gait classification method based on the deep belief network (DBN) optimized by the sparrow search algorithm (SSA) is proposed. The multiple features obtained based on surface electromyography (sEMG) are fused. These functions are used to train the model. First, the sample features, such as the time domain and frequency domain features of the denoised sEMG are extracted and then the fused features are obtained by feature combination. Second, the SSA is utilized to optimize the architecture of DBN and its weight parameters. Finally, the optimized DBN classifier is trained and used for gait recognition. The classification results are obtained by varying different factors and the recognition rate is compared with the previous classification algorithms. The results show that the recognition rate of SSA-DBN is higher than other classifiers, and the recognition accuracy is improved by about 2% compared with the unoptimized DBN. This indicates that for the application in gait recognition, SSA can optimize the network performance of DBN, thus improving the classification accuracy.

Analysis of hydraulic system’s weak fault based on variable parameter multi-scale permutation entropy and deep belief network

Aiming at automatic recognition of hydraulic system’s weak fault, an analysis method based on variable parameter multi-scale permutation entropy (VPMPE) and deep belief network (DBN) is proposed in this paper. The external vibration signals of experimental hydraulic equipment in normal state and three different leakage state (slight, moderate and severe) are taken as research object. By the proposed method, experiment signals are first processed to obtain their multi-scale permutation entropy in different conditions by changing the embedding dimension m and the scale factor s, then the multi-scale permutation entropy under different m and s are combined to form feature vectors, and lastly the DBN classifier is used to identify and analyze the testing samples. Verification test shows that the proposed method has good effects on hydraulic system’s weak fault, which can accurately judge whether there is leakage fault and measure the fault severity.

Online product sentiment analysis using random evolutionary whale optimization algorithm and deep belief network

• Image and text normalization techniques are used to improve the data quality. • Used MLBP , SURF, TF-IDF and LSA techniques for feature extraction. • Used RE-WOA to reduce the dimension of the extracted feature vectors. • Used DBN classifier to classify the user's sentiments of the online products. In the recent decades, the online product sentiment analysis is an emerging research topic that assists the customers to take better decisions on purchasing the products and to achieve better sales of the products. Recently, several machine learning techniques are experimented on many datasets for analyzing the customer's sentiments through online portals. Still, the customers are struggling to obtain the aspect sentiments expressed by other customers, particularly in the amazon websites. Therefore, a novel automated model is proposed in this manuscript for an effective online product sentiment analysis. After collecting the multimodal data from the amazon websites, the image and data normalization techniques are employed for better understanding of the collected data. Further, the feature extraction is performed by utilizing Latent Semantic Analysis (LSA), Term Frequency- Inverse Document Frequency (TF-IDF), Modified Local Binary Pattern (MLBP), and Speeded Up Robust Features (SURF) descriptors for extracting the textual and visual feature vectors from the preprocessed data. Finally, the Random Evolutionary Whale Optimization Algorithm (REWOA) and Deep Belief Network (DBN) classifier are integrated for feature vector optimization and sentiment classification. By using feature optimization, the system complexity and running time of the classifier is improved. The experimental investigation states that the developed REWOA-DBN model achieved 96.86% of classification accuracy, which is better compared to other optimizers and classifiers

Security framework for IoT and deep belief network-based healthcare system using blockchain technology

Nowadays, the computerisation and information systems need very fast, easier and secure data analysis models. So, blockchain technology and deep learning concepts play important roles for data security and analysis in field of healthcare applications. Blockchain technology has optimal reliability in various sectors like healthcare, security, information management, smart home, etc. In this paper, blockchain technology is used for the healthcare system to provide a secure and reliable data transmission. Furthermore, IoT-based wearable devices are used to collect the data from the patients. Moreover, deep belief network (DBN) is proposed to classify the obtained data in order to predict the type of disease or problems of the patients. The experimental results shows that the proposed framework and classification model abruptly outperforms when compared with other existing techniques. The efficiency of the proposed framework is analysed based on the performance metrics like throughput and latency. The obtained loss ratio for the proposed model seems very less of 0.0126 when compared with other existing techniques like MLP, SVM, ANN, and CNN. The accuracy obtained for the proposed DBN classifier is above 95% for the increase in the number of patients.

Predicting Cardiovascular and Cerebrovascular Events Based on Instantaneous High-Order Singular Entropy and Deep Belief Network.

Automatically predicting cardiovascular and cerebrovascular events (CCEs) is a key technology that can prevent deaths and disabilities. Herein, we propose predicting CCE occurrences based on heart rate variability (HRV) analysis and a deep belief network (DBN). The proposed prediction algorithm uses eight novel HRV signal features, which are calculated based on the following steps. First, the instantaneous amplitude (IA), instantaneous frequency (IF), and instantaneous phase (IP) are calculated for the HRV signals. Second, the high-order cumulant is estimated for the HRV and its IA, IF, and IP. Third, a high-order singular entropy is calculated to measure the fluctuation in signals. Fourth, eight novel features are obtained and processed using a DBN classifier designed for CCE prediction. The DBN classification method, with the novel HRV features, outperformed existing methods in terms of accuracy. Thus, the scheme proposed herein provided a novel direction for predicting CCEs.

Breast cancer detection in digital mammography using a novel hybrid approach of Salp Swarm and Cuckoo Search algorithm with deep belief network classifier

ABSTRACT The Computer Aided Diagnosis (CAD) system has evolved as a useful tool for radiologists to classify breast cancer images into various categories, enabling early diagnosis and treatment. In CAD model construction, feature selection is essential for determining a subset of appropriate features to diagnose breast cancer. Salp Swarm Algorithm (SSA) is an evolutionary algorithm that simulates the swarming behaviour of salps. SSA has some advantages such as simplicity, speed in searching and ease of hybridization with other optimization algorithms. However, it suffers from being stuck in local optima and having slow convergence. To address these issues, this work proposes a novel hybridization algorithm called SSACS by combining the SSA with Cuckoo Search (CS) to improve convergence and exploitation capabilities. Further, the Deep Belief Network (DBN) classifier is applied to classify the mammogram images and improve the diagnosis rates. The proposed system’s efficacy is validated with the benchmark database of the Mammographic Image Analysis Society (mini-MIAS) dataset. The experimental findings indicate that the proposed SSACS with DBN classifier outperforms the state-of-the-art methods.

Improved Grasshopper Optimization Algorithm based Feature Selection with Evolutionary Outlay-Aware Deep Belief Network Classifier (IGOA-EOA-DBNC) for High Dimensional Datasets

Background: High dimensional datasets contain the curse of dimensionality, and hence data mining becomes a more difficult task. Feature selection in the knowledge data and discovery process provides a solution for this curse of dimensionality issue and helps the classification task reduce the time complexity and improve the accuracy. Objectives: This paper aims to recognize a bio-inspired algorithm that best suits feature selection and utilizes optimized feature selection techniques. This algorithm is used to design machine learning classifiers that are suitable for multiple datasets and for both high dimensional datasets, moreover to carry out performance analysis with regards to the accuracy of a classification and the processing time for classification. Methods: This study employs an improved form of grasshopper optimization algorithm to perform feature selection task. Evolutionary outlay aware deep belief network is used to perform the classification task. Findings: In this research, 20 UCI benchmark data sets are taken with full 60 features and 30000 instances. The datasets are Mammography, Monks-1, Bupa, Credit, Parkinson's, Monk-2, Sonar, Ecoli, Prognostic, Ionosphere, Monk-3, Yeast, Car, Blood, Pima, Spect, Vert, Prognostic, Contraceptive, and Tic-Tac-Toe endgame. Table 1 describes the dataset details, number of instances, datasets and features. The overall performance is performed using MATLAB 6.0 tool, which runs on Microsoft Windows 8, and the configuration is Core 13 processor with 1 TB hard disk and 8GB RAM. Performance standards, like classification accuracy and the processing time for classification, is achieved. Novelty: Interestingly, the Improved Grasshopper Optimization Algorithm uses error rate and classification accuracy of the Evolutionary Outlay Aware –Deep Belief Network Classifier as fitness function values. This combined work of classification and feature selection is briefly represented as IGOA-EOA-DBNC. Twenty datasets are selected for testing the performance regarding elapsed time and accuracy, which gives better results.

Deep Appearance Model and Crow-Sine Cosine Algorithm-Based Deep Belief Network for Age Estimation

Age estimation has been paid great attention in the field of intelligent surveillance, face recognition, biometrics, etc. In contrast to other facial variations, aging variation presents several unique characteristics, which make age estimation very challenging. The overall process of age estimation is performed using three important steps. In the first step, the pre-processing is performed from the input image based on Viola-Jones algorithm to detect the face region. In the second step, feature extraction is done based on three important features such as local transform directional pattern (LTDP), active appearance model (AAM), and the new feature, deep appearance model (Deep AM). After feature extraction, the classification is carried out based on the extracted features using deep belief network (DBN), where the DBN classifier is trained optimally using the proposed learning algorithm named as crow-sine cosine algorithm (CS).

Epilson Swarm Optimized Cluster Gradient and deep belief classifier for multi-attack intrusion detection in MANET

Design of intrusion detection, and MANET prevention mechanism, with scrutinized detection rate, memory consumption with minimal overhead are crucial research concerns. Node mobility and energy of the node are dual essential optimization issues in mobile ad hoc networks (MANETs) where nodes traverse uncertainly in any direction, evolving in topology's continuing modification. A Centrality Epilson Greedy Swarm and Gradient Deep Belief Classifier (CEGS-GDBC) for multi-attack intrusion detection are designed with the proposed method. The paper concentrates on the issues of node mobility and energy to emerge a clustering algorithm inspired by Dual Network Centrality for cluster head election in MANET. Compact cluster formation is done with the help of Epilson Greedy Swarm Optimization. Finally, with a hybrid type of IDS, Gradient using the Deep Belief Network Classifier identifies multi-attack, i.e., DoS and Zero-Day attack. The proposed work is experimented extensively in the NS-2 network simulator and compared with the other existing algorithms. The proposed method's performance is studied in terms of different parameters such as attack detection rate, memory consumption, and computational time for identifying and isolating the intruder. Simulation results show that the proposed method extensively minimizes the IDS traffic and overall memory consumption and maintains a high attack detection rate with minimal computational time. From the results, CEGS-GDBC method increases the attack detection rate by 31% and reduces the memory consumption and computational time by 39% and 41% as compared to Fuzzy elephant—Herd optimization and Cross centric intrusion detection system.

Human Activity Recognition based on Deep Belief Network Classifier and Combination of Local and Global Features

During the past decades, recognition of human activities has attracted the attention of numerous researches due to its outstanding applications including smart houses, health-care and monitoring the private and public places. Applying to the video frames, this paper proposes a hybrid method which combines the features extracted from the images using the ‘scale-invariant features transform’ (SIFT), ‘histogram of oriented gradient’ (HOG) and ‘global invariant features transform’ (GIST) descriptors and classifies the activities by means of the deep belief network (DBN). First, in order to avoid ineffective features, a pre-processing course is performed on any image in the dataset. Then, the mentioned descriptors extract several features from the image. Due to the problems of working with a large number of features, a small and distinguishing feature set is produced using the bag of words (BoW) technique. Finally, these reduced features are given to a deep belief network in order to recognize the human activities. Comparing the simulation results of the proposed approach with some other existing methods applied to the standard PASCAL VOC Challenge 2010 database with nine different activities demonstrates an improvement in the accuracy, precision and recall measures (reaching 96.39%, 85.77% and 86.72% respectively) for the approach of this work with respect to the other compared ones in the human activity recognition.

Local gradient pattern and deep learning-based approach for the iris recognition at-a-distance

One of the biometric techniques utilized to predict the human is based on the iris. The recognition of iris is performed by discovering an individual without human intervention utilizing the iris of human eyes. Iris offers distinct information about the person. This research presents deep learning strategy for performing iris recognition. Primarily, image is pre-processed to obtain exact region iris. Then, region of iris is extracted using Hough Transform, followed by segmentation and normalization of iris region using Daugman’s rubber sheet model. Once segmentation is performed, the features are generated with ScaT-LOOP that is the combination of Scattering Transform (ST), Tetrolet transforms (TT), Local Gradient Pattern (LGP) and Local Optimal Oriented Pattern (LOOP). Finally, steepest gradient-based Deep Belief Network (DBN) is utilized for recognizing the iris. The performance of iris recognition using the DBN classifier is computed based on accuracy, False Rejection Rate (FRR), and False Acceptance Rate (FAR). The proposed method achieves maximum accuracy of 97.96%, minimal FAR of 0.493%, and minimal FRR of 0.48% that indicates its superiority.

Modified PLVP with Optimised Deep Learning for Morphological based Road Extraction

ABSTRACT This paper introduces a new modified local pattern descriptor to extract road from rural areas’ aerial imagery. The introduced local pattern descriptor is actually the modification of the proposed local vector pattern (P-LVP), and it is named as Modified-PLVP (M-PLVP). In fact, M-PLVP extracts the texture features from both road and non-road pixels. The features are subjected to train the Deep belief Network (DBN); thereby the unknown aerial imagery is classified into road and non-road pixel. Further, to improve the classification rate of DBN, morphological operations and grey thresholding operations are performed and so that the road segmentation is performed. Apart from this improvement, this paper incorporates the optimisation concept in the DBN classifier, where the activation function and the count of hidden neurons are optimally selected by a new Trail-based WOA (T-WOA) algorithm, which is the improvement of the Whale Optimisation Algorithm (WOA). Finally, the performance of proposed M-PLVP is compared over other local pattern descriptors concerning measures like Accuracy, Sensitivity, Specificity, Precision, Negative Predictive Value (NPV), F1Score and Mathews correlation coefficient (MCC), False positive rate (FPR), False negative rate (FNR), and False Discovery Rate (FDR), and proves the betterments of M-PLVP over others.

Violation Detection in Service Level Agreement to Ensure the Privacy in Cloud Community using Chicken Spider Monkey Optimization-Based Deep Belief Network

In the cloud service system, the expectation between the customer and the cloud provider for the delivered services like, responsibilities and penalties of the violation arise in the cloud parties are specifically designed in a document named as Service Level Agreement (SLA). However, violation in the Service Level Agreement is a major challenge in the research community of SaaS cloud providers. Hence, an effective violation detection method named Chicken Spider Monkey Optimization-based Deep Belief Network (ChicSMO-based DBN) is proposed in this research to perform the violation detection mechanism in the cloud environment. The proposed Chicken Spider Monkey Optimization-based Deep Belief Network effectively computes the scores, like User Privacy Score (UPS), Service Level Agreement score (SLA), and Cloud Service Provider score (CSP) based on the fitness value. The spider monkeys forage the food by evaluating the distance from the food, and update their position towards the food source. The feature scores are computed by updating the position of local and global leader group members, and perform the violation detection process using the Deep Belief Network classifier. The performance enhancement attained by the proposed algorithm in terms of the evaluation metrics, like accuracy, coverage rate, and F-value which acquire with the values of 0.862, 0.944, and 0.889 with the transaction data of 1000, respectively. It increases the reliability to address the QoS guarantee issues and increase the confidence level of the prediction model.

Design and Application of Deep Belief Network Based on Stochastic Adaptive Particle Swarm Optimization

Due to the problem of poor recognition of data with deep fault attribute in the case of traditional superficial network under semisupervised and weak labeling, a deep belief network (DBN) was proposed for deep fault detection. Due to the problems of deep belief network (DBN) network structure and training parameter selection, a stochastic adaptive particle swarm optimization (RSAPSO) algorithm was proposed in this study to optimize the DBN. A stochastic criterion was proposed in this method to make the particles jump out of the original position search with a certain probability and reduce the probability of falling into the local optimum. The RSAPSO-DBN method used sample data to train the DBN and used the final diagnostic error rate to construct the fitness value function of the particle swarm algorithm. By comparing the minimum fitness value of each particle to determine the advantages and disadvantages of the model, the corresponding minimum fitness value was selected. Using the number of network nodes, learning rate, and momentum parameters, the optimal DBN classifier was generated for fault diagnosis. Finally, the validity of the method was verified by bearing data from Case Western Reserve University in the United States and data collected in the laboratory. Comparing BP (BP neural network), support vector machine, and heterogeneous particle swarm optimization DBN methods, the proposed method demonstrated the highest recognition rates of 87.75% and 93.75%. This proves that the proposed method possesses universality in fault diagnosis and provides new ideas for data identification with different fault depth attributes.

Chicken swarm foraging algorithm for big data classification using the deep belief network classifier

PurposeThe innovation in big data is increasing day by day in such a way that the conventional software tools face several problems in managing the big data. Moreover, the occurrence of the imbalance data in the massive data sets is a major constraint to the research industry.Design/methodology/approachThe purpose of the paper is to introduce a big data classification technique using the MapReduce framework based on an optimization algorithm. The big data classification is enabled using the MapReduce framework, which utilizes the proposed optimization algorithm, named chicken-based bacterial foraging (CBF) algorithm. The proposed algorithm is generated by integrating the bacterial foraging optimization (BFO) algorithm with the cat swarm optimization (CSO) algorithm. The proposed model executes the process in two stages, namely, training and testing phases. In the training phase, the big data that is produced from different distributed sources is subjected to parallel processing using the mappers in the mapper phase, which perform the preprocessing and feature selection based on the proposed CBF algorithm. The preprocessing step eliminates the redundant and inconsistent data, whereas the feature section step is done on the preprocessed data for extracting the significant features from the data, to provide improved classification accuracy. The selected features are fed into the reducer for data classification using the deep belief network (DBN) classifier, which is trained using the proposed CBF algorithm such that the data are classified into various classes, and finally, at the end of the training process, the individual reducers present the trained models. Thus, the incremental data are handled effectively based on the training model in the training phase. In the testing phase, the incremental data are taken and split into different subsets and fed into the different mappers for the classification. Each mapper contains a trained model which is obtained from the training phase. The trained model is utilized for classifying the incremental data. After classification, the output obtained from each mapper is fused and fed into the reducer for the classification.FindingsThe maximum accuracy and Jaccard coefficient are obtained using the epileptic seizure recognition database. The proposed CBF-DBN produces a maximal accuracy value of 91.129%, whereas the accuracy values of the existing neural network (NN), DBN, naive Bayes classifier-term frequency–inverse document frequency (NBC-TFIDF) are 82.894%, 86.184% and 86.512%, respectively. The Jaccard coefficient of the proposed CBF-DBN produces a maximal Jaccard coefficient value of 88.928%, whereas the Jaccard coefficient values of the existing NN, DBN, NBC-TFIDF are 75.891%, 79.850% and 81.103%, respectively.Originality/valueIn this paper, a big data classification method is proposed for categorizing massive data sets for meeting the constraints of huge data. The big data classification is performed on the MapReduce framework based on training and testing phases in such a way that the data are handled in parallel at the same time. In the training phase, the big data is obtained and partitioned into different subsets of data and fed into the mapper. In the mapper, the features extraction step is performed for extracting the significant features. The obtained features are subjected to the reducers for classifying the data using the obtained features. The DBN classifier is utilized for the classification wherein the DBN is trained using the proposed CBF algorithm. The trained model is obtained as an output after the classification. In the testing phase, the incremental data are considered for the classification. New data are first split into subsets and fed into the mapper for classification. The trained models obtained from the training phase are used for the classification. The classified results from each mapper are fused and fed into the reducer for the classification of big data.

Detection of distributed denial of service attack in cloud computing using the optimization-based deep networks

ABSTRACT Cloud computing services provide a wide range of resource pool for maintaining a large amount of data. Cloud services are commonly used as the private or public data forum based on the demand, and the increase in usage has lead to security concerns. The information in the cloud comes under threat due to hackers, and the most common attack on the cloud data is considered as the Distributed Denial of Service (DDoS) attack. This work has concentrated on detecting the DDoS attack by developing the deep learning-based classifier. The service request from the users is collected and grouped as the log information. From the log file, some important features are selected for the classification using the Bhattacharya distance measure to reduce the training time of the classifier. Here, Taylor-Elephant Herd Optimisation based Deep Belief Network (TEHO-DBN), is developed by modifying the Elephant Herd Optimisation (EHO) with the Taylor series and the algorithm thus developed is adopted to train the Deep Belief Network (DBN) for the DDoS attack detection. From the simulation results, it can be concluded that the proposed TEHO based DBN classifier has improved performance with a maximum accuracy of 0.830.

Character Recognition Algorithm Based on Fusion Probability Model and Deep Learning

Abstract The authors proposed a character recognition algorithm to improve the recognition efficiency and recognition accuracy in character recognition. The algorithm is based on a deep belief network classifier. In the character recognition, a comprehensive feature model is established firstly by combining the histogram Gabor feature, grid level feature and gray level co-occurrence matrix. Subsequently, the deep belief network trains the feature model. Finally, the probability model is used to judge the recognition symbols. The algorithm is tested with 74 k data set and is compared and analyzed from three indexes: false acceptance rate, false rejection rate and accuracy. The data set simulation and comparison with other algorithms indicate that the recognition system based on the probability model and depth learning has higher accuracy and better performance.