Biomimetic Pattern Recognition Research Articles

Aviation application of dual-weighted neural network based on biomimetic pattern recognition

Abstract In order to solve the difficult problem of fault diagnosis in aviation equipment, this paper uses a biomimetic pattern recognition method. Compared to traditional pattern recognition, biomimetic pattern recognition is able to construct closed, complex geometries to cover all kinds of samples. Biomimetic pattern recognition is a “cognitive” approach, where the process between two similar things is considered asymptotic. It mathematically means that all similar samples in the feature space are continuously indistinguishable. The paper establishes a dual-weighted neural network model of an aircraft auxiliary power unit. Comparing the diagnostic process and results with traditional manual neural networks. The results show that biomimetic pattern recognition is a feasible and efficient diagnostic method for the fault diagnosis of aviation equipment. It provides a new way of thinking and method for the practical application of aviation equipment fault diagnosis.

Multiple Microphone Speaker Recognition System for Second Language Based on Biomimetic Pattern Recognition with Big Data Fusion

The speaker recognition problem for second language based on biomimetic pattern recognition with big data fusion by multiple microphone has been addressed in this paper. Biomimetic pattern recognition is a new machine learning algorithm which can be used to study the geometric characteristics of a large number of sample points in high-dimensional space. Machine learning is an important way for computer to realize intelligence. The development of artificial intelligence can not be separated from the support of machine learning. Big data Unifield architecture of Hadoop system integrates machine learning and big data processing. Several speaker feature extraction methods of big data are described. Cepstrum and Δcepstrum can offer a significant computational advantage in the eigenvalue problem in result of enhancing correct rate. The big data fusion framework of neural network multiple microphone verification system is presented; in particular, a neural network constructed by multiple value neural network algorithm of big data fusion is also proposed for adjusting the parameters. The experimental results of big data fusion using multiple microphone speaker recognition system for second language illustrate the effectiveness of the proposed biomimetic pattern recognition method.

Geometric Cognition Method of Biomimetic Pattern Recognition and Its Application in the Digitalization of Modern Guangdong Customs Archives

This paper puts forward a digitalization model of Modern Guangdong Customs archives based on biomimetic pattern recognition. By introducing biomimetic pattern recognition, the digitalization of Modern Guangdong Customs archives becomes a social production of information network. With this model, we can not only make full use of the wisdom of biomimetic pattern recognition, but also digitalize the Modern Guangdong Customs archives with high efficiency and low cost, thus fully exploring the internal value of these archives. In this paper, firstly, we propose Neuron Model of Digital Biomimetics, and investigate Geometric Sequential Learning Ahead Mask in digitalizing Modern Guangdong Customs archives. Then, we analyze the application of digital biomimetics to the digitalization of Modern Guangdong Customs archives. Finally, we analyze the of comparison of the work cost among Taobao, Professional Institutions, Crowdsourcing and Digital Biomimetics, which proves that the geometric neuron model is reasonable and effective.

Image Classification Using Biomimetic Pattern Recognition with Convolutional Neural Networks Features.

As a typical deep-learning model, Convolutional Neural Networks (CNNs) can be exploited to automatically extract features from images using the hierarchical structure inspired by mammalian visual system. For image classification tasks, traditional CNN models employ the softmax function for classification. However, owing to the limited capacity of the softmax function, there are some shortcomings of traditional CNN models in image classification. To deal with this problem, a new method combining Biomimetic Pattern Recognition (BPR) with CNNs is proposed for image classification. BPR performs class recognition by a union of geometrical cover sets in a high-dimensional feature space and therefore can overcome some disadvantages of traditional pattern recognition. The proposed method is evaluated on three famous image classification benchmarks, that is, MNIST, AR, and CIFAR-10. The classification accuracies of the proposed method for the three datasets are 99.01%, 98.40%, and 87.11%, respectively, which are much higher in comparison with the other four methods in most cases.

Study of Modeling Samples Selection Method Based on Near Infrared Spectrum

For more wheat varieties classification problem, we use near infrared spectrumto do qualitative analysis. Increasing the size of modeling sample could increase information of the model, however, at the same time, it also makes information redundancy so that modeling time and storage space will increase, thus, we need to decrease the size of modeling sample though selecting them. Some information must be lost and the effects of the model must be worse if we select samples blindly. We put forward the k nearest neighbor-density sample selection based on the traditional selection methods. Experiments use the near infrared diffuse reflection spectrum of wheat seed from lots of days. First, we use preprocessing and feature extraction to deal with the wheat original spectrum, then select modeling sample by three methods that are random sampling, k nearest neighbor and k nearest neighbor-density, finally, we establish the models of BPR(Biomimetic Pattern Recognition) and BPRI(Biomimetic Pattern Recognition Improved). The experimental results show that in the model of BPR we get the best results using the selection method of k nearest neighbor-density, especially it also decreases the size of modeling sample deeply, and in the model of BPRI the results using the selection method of k nearest neighbor-density are much better than random sampling and a little better than k nearest neighbor, but in the meanwhile the size of modeling sample using the selection method of k nearest neighbor-density are much smaller than k nearest neighbor. The experimental results prove that the sample selection method of k nearest neighbor-density can not only greatly reduce the modeling sample size, and ensure the quality of the model, it has obvious effect on varieties classification problem of wheat.

Feasibility of analyzing frost-damaged and non-viable maize kernels based on near infrared spectroscopy and chemometrics

The current detection methods for maize seed damage and viability assessment are usually time-consuming, tedious, and costly. In this investigation, near infrared spectroscopy (NIRS), a nondestructive and rapid analytical method, was explored to analyze frost-damaged and non-viable seeds. Principal component analysis (PCA), partial least squares (PLS) and orthogonal linear discriminant analysis (OLDA) were combined to extract feature from near infrared (NIR) spectra. PLS + OLDA can extract difference characteristics of normal and frost-damaged seeds more efficiently than PCA + OLDA. Three classification algorithms were utilized and compared: Support vector machine (SVM), biomimetic pattern recognition (BPR), and mahalanobis distance (MD). BPR can classify normal and frost-damaged seeds better and achieved the highest average accuracy of 97%. Discrimination model of viable and non-viable seeds in frost-damaged seeds based on SVM, BPR and MD were built and achieved accuracies of 94%, 97.25% and 89.5% respectively. BPR model yielded the most correct value in predicting germination rate of validation set, and improved the germination rate of validation set dramatically from 27.5% to 100% by screening out non-viable seeds. NIRS and chemometrics as demonstrated in this paper can provide a novel method which can assess and improve quality of maize seed quickly and inexpensively.

Edge Detection Based on Biomimetic Pattern Recognition

This paper introduced a technique for edge detection by Biomimetic pattern recognition (BPR). The images were scanned into a computer. Due to the nature of the acquiring technique, the acquired images have lots of artifacts, resulting in complicated edge detection. According to this, we used Biomimetic pattern recognition, which is based on “matter cognition” instead of “matter classification” and rather closer to the function of the human being. Finally, the experiments showed that the technique is feasible and has some flexibility.

Variety identification method of coated maize seeds based on near-infrared spectroscopy and chemometrics

Near infrared spectroscopy (NIRS) has been used to identify variety authenticity of bare maize kernels. However, maize seeds are coated with seed coating agents in practice. Therefore it is of great significance to investigate the feasibility of identifying coated maize kernels by NIRS. This study employed NIRS to quickly determine the variety to which a sample of coated seed belonged. The NIR spectra of clean and coated seeds were obtained using MPA spectrometry in diffuse reflectance mode by three methods. Influence of seed coating agent on NIR spectra was discussed based on spectra collected from whole single seed by method 1. In method 2, to eliminate the influence of seed coat, the seed coat agent on the back side of the seed was polished away. In method 3, clean and coated seeds were cut open along the crease on the embryo side, and the sections were scanned by the spectrometer, so as to acquire information on the seed, and to avoid the influence of the seed coating agent. Then, support vector machine (SVM), soft independent modeling of class analogy (SIMCA), and biomimetic pattern recognition (BPR) were employed to establish the identification model for four maize varieties. Performance of variety models based on spectra measured by method 1 and 2 were poor. In method 3, the SIMCA model showed better performance than SVM and BPR models, and achieved an accuracy rate of 97.5%. The results demonstrated the feasibility of utilizing NIRS and chemometrics, as an objective and rapid method for the identification of maize seeds with seed coating agents.

Biomimetic Pattern Recognition for Classification of Proteomic Profile

We propose a biomimetic pattern recognition (BPR) approach for classification of proteomic profile. The proposed approach preprocess profile using iterative minimum in adaptive setting window (IMASW) method for baseline correction, discrete wavelet transform (DWT) for fitting and smoothing, and average total ion normalization (ATIN) for remove the influence of vary amount of sample and degradation over time. Then principal component analysis (PCA) and BPR build classification model. With an optimization of the parameters involved in the modeling, we obtain a satisfactory model for cancer diagnosis in three proteomic profile datasets. The predicted results show that BPR technique is more reliable and efficient than support vector machine (SVM) method.

A co-training algorithm for EEG classification with biomimetic pattern recognition and sparse representation

In this paper, a semi-supervised co-training method is proposed to construct a high-performance classifier for motor imagery based Brain–Computer Interfaces (BCI), which is trained with little labeled data. Two novel classifiers: Biomimetic Pattern Recognition (BPR) and Sparse Representation (SR) are applied. Each classifier labels the unlabeled data for the other classifier to extend the labeled set. And the enlarged labeled set is used to generate the final classifier BPR–SR. BPR–SR is constructed by combining BPR with SR, in which SR is used to replace the traditional distance method when BPR encounters the overlap coverage problem. The features extracted by common spatial pattern were used for classification. The experiment with datasets from previous BCI competitions demonstrates the validity of our proposed co-training algorithm compared with BPR, SR and BPR–SR without co-training. And the experiment about our novel classifier BPR–SR compared with other excellent classifiers on datasets from previous BCI competitions and our own lab was also conducted. And the result shows that the classifier BPR–SR can overcome the overlap coverage problem of BPR by introducing SR.

A Novel SAR Image Recognition Algorithm with Rejection Mode via Biomimetic Pattern Recognition

Due to increasing demand of the modern battlefield awareness and military reconnaissance, Synthetic Aperture Radar (SAR) Automatic Target Recognition (ATR) has been receiving more and more attention. In this paper, a SAR ATR algorithm with rejection mode based on Local Phase Quantization (LPQ) plus Biomimetic Pattern Recognition (BPR) has been proposed. There are three main steps in the proposed algorithm: firstly, a simple preprocessing procedure based on centroid location is applied to the original SAR image to extract the target’s Region of Interest (ROI). Secondly, Short-term Fourier Transform (STFT) computed over sub-windows at every pixel of the ROI image is performed and the extracted information of the Fourier phase spectrum is quantized as features to overcome the target’s azimuth angle variations. Finally, high dimensional geometry space covering method based on BPR theory is adopted and hyper-sausage neuron is employed to construct geometry coverage for the purpose of recognizing or rejecting the incoming targets. Experiments on the public standard MSTAR database show that this proposed algorithm can achieve relatively high recognition rate while obtaining high rejection rate for unknown incoming targets.

A MULTI-SCALE LOCAL PHASE QUANTIZATION PLUS BIOMIMETIC PATTERN RECOGNITION METHOD FOR SAR AUTOMATIC TARGET RECOGNITION

Synthetic aperture radar (SAR) automatic target recog- nition (ATR) has been receiving more and more attention in the past two decades. But the problem of how to overcome SAR target am- biguities and azimuth angle variations has still left unsolved. In this paper, a multi-scale local phase quantization plus biomimetic pattern recognition (BPR) method is presented to solve these two di-culties. By applying multiple scales local phase quantization (LPQ) on the ob- served SAR images, the blur and azimuth invariant features can be extracted, and these features are fusion at consecutive multiple scales to achieve better results. Then PCA method is applied to further re- duce the feature dimension and achieve its e-ciency. Finally, high dimensional space geometry covering method based on BPR theory is adopted to construct hyper sausage neuron links for target recognition. Experiments on the MSTAR database show that the proposed method can achieve satisfying recognition accuracy compared with other state- of-the-art methods.

Constructive Algorithm of Multi-Weight Neuron Covering Networks

Biomimetic Pattern Recognition theory uses multi-weight neuron covering networks to perform recognition tasks.However,there are few discussions on the constructive methods of multi-weight neuron covering networks,namely how many neurons are needed to fulfill the covering of a data set.A large number of neurons can cover the data set precisely,but it will make the network be more complicated.In this paper,we propose a constructive algorithm for multi-weight neuron covering networks.The algorithm can construct a covering network with least neurons and maintain the covering ability of the networks simultaneously.The experiments show the efficiency of the proposed algorithm.

A novel method for motor imagery EEG adaptive classification based biomimetic pattern recognition

The real on-line BCI is indeed a hotspot at present whose performance however is limited by the problems of non-stationary etc. In this paper, a novel method for the adaptive classification of motor imagery EEG data based Biomimetic Pattern Recognition (BPR) through introducing three adaptive operators is proposed. Considering that the large amounts of labeled samples are difficult to get in the actual application, we also propose a novel unsupervised scheme based the adaptive classifier to solve this problem. Sufficient experiments are conducted on the datasets from previous Brain–Computer Interface Competitions and the actual on-line EEG data in adaptive scheme. The results demonstrate that the new algorithm is efficiency and robust compared with non-adaptive classifiers. Besides, a couple of analyses are made on the selection of parameters in the adaptive BPR, and some advice has been come up with about their selections.

Non-Specific Person Continuous Speech Identification in Second Language using BPR

Second language speech recognition is an important technical means of man-machine communication system. In this paper, we propose a biomimetic pattern recognition (BPR) algorithm for non specific person continuous speech identification in second language. Feature parameters are extracted directly from single number sample being segmented according to Mel cepstral way. BPR-based connected number recognition experiments show that the proposed method with better identification performance in second language than NN neural network and SVM. From non specific person continuous speech recognition experiments, we show that the BPR algorithm greatly improves word recognition accuracy and has good recognition ability without causing poorer performance in second language. DOI: http://dx.doi.org/10.11591/telkomnika.v10i7.1550 Full Text: PDF

Biomimetic Pattern Recognition Method for Breast Cancer Using Gene Expression Data

The purpose of this study was to develop a novel prediction method for breast cancer based on gene expression data through using a susceptible marker-selectable biomimetic pattern recognition (BPR) method, with which a parameter increasing method (PIM) was proposed to incorporate. The method was used to predict early detection, transition from normal cell to cancerous cell and prognosis signature of patients with adjuvant systemic therapy. Several genes were selected as susceptible genes associated with breast cancer. It can be shown by the results that the “cognition” BPR method could correctly predict detection, cancerous cell transition and good or poor prognosis signature with approximate 85%, 98% and 88% accuracy separately. In order to study the performance of BPR, Fisher discriminant analysis (FDA) and support vector machine (SVM) methods also were applied to analyze the gene expression data. From the results, it can be found that the BPR method is superior to FDA and SVM with respect to classification ability. Furthermore, the prediction performance can be improved through using biomarker instead of whole gene expression data for any method.

A New Method of Signature Verification Based on Biomimetic Pattern Recognition Theory

Aim at the difficulty and low recognition rate of signature verification, this paper introduces biomimetic pattern recognition theory and applies it to the problem. According to the features of the signature samples, the coverage in the high-dimension feature space is built, one class of samples are all covered with a super-sausage neuron chain. As the radius selection of the super-sausage neurons maybe unreasonable, unwanted area may be covered and correct recognition rate will reduce. So this paper uses the relationship of the distance between the two training samples and the average distance of all the neurons to adjust the radius of the super-sausage neuron automatically. Finally, the experiments show that compared to traditional pattern recognition method, biomimetic pattern recognition theory used in signature verification have a better recognition result and is more effective.

Tone recognition based on biomimetic pattern recognition theory

Pitch frequency trajectories can more truly reflect the tone characteristics of Mandarin.It is a better way to identify tone by identifying different pitch trajectories.According to the theory of biomimetic pattern recognition,an improved tone recognition algorithm was proposed.It used Iterative Self-Organizing Data Analysis Techniques Algorithm (ISODATA) to find centers of coverage areas,and multi-weight neural network to achieve coverage for each cluster center.The experimental results show that in the case of a small sample,the proposed algorithm can get a higher recognition rate than that of Hidden Markov model (HMM) and Support Vector Machine (SVM).

DIVISION OF SPACE BASED SEGMENTATION SYSTEM OF UNCONSTRAINED NUMERAL STRINGS

This paper presents a new segmentation method based on division of space which chooses principal curves to extract strokes of characters and form the initial stroke set. The strokes in the initial set are disposed by the fuzzy theorems and grouped based on the confidence of the classifiers. The training sample space is constructed by the Affinity Propagation (AP) algorithm and the Biomimetic Pattern Recognition (BPR) theory. The strokes are arranged in a sequence divided by calculating the distance of each subsequence to the relative subspace. Finally, with the maximum posterior (MAP) criterion, the optimal segmentation hypotheses described in a probabilistic model is found. Our system is validated by experimental results from samples of real Chinese bank checks, achieving a correct rate of without 95.23% rejection.

A retrieval method for human Mocap data based on biomimetic pattern recognition

A retrieval method for human Mocap (Motion Capture) data based on biomimetic pattern recognition is presented in this paper. BVH rotation channels are extracted as features of motion for both the retrieval instance and the motion data. Several hyper sausage neurons are constructed according to the retrieval instance, and the trained domain covered by these hyper sausage neurons can be considered as the distribution range of a same kind of motions. By use of CMU free motion database, the retrieval algorithm has been implemented and examined, and the experimental results are illustrated. At the same time, the main contributions and limitations are discussed.