Papers by Keyword: Probabilistic Neural Network (PNN)

Abstract: This paper uses the probabilistic neural network (PNN) to monitor the operation statues for the compressor of air-conditioners. The field data including the high/low pressures and the high/low temperatures of refrigerants are measured in a practical system. PNN analyses the refrigerants’ pressures/temperatures of air-conditioners to monitor the operation conditions of compressor. PNN method is suitable for application in a dynamic environment by using new data-set and new hidden without doing any computed iteration. Computer simulations were conducted with refrigerants’ records, test results showed the effectiveness of the proposed system.

Abstract: To solve the difficulties of establishing precise mathematical model of ball bearing fault diagnosis, a classification method based on probabilistic neural network (PNN) used for ball bearing fault mode classification is proposed. Firstly, this paper analyzed the basic theory of PNN, and then a mapping relationship between feature vector and fault mode is set up based on PNN. Secondly, selections of ball bearing fault features and practical procedure of neural network setting and training are discussed. Experiments and compared with the algorithm of back propagation neural network (BPNN) prove that PNN method is feasible and has better diagnosis efficiency than BPNN.

Abstract: The Hearing Impaired People (HIP) cannot distinguish the sound from a moving vehicle approaching from their behind. Since, it is difficult for hearing impaired to hear and judge sound information and they often encounter risky situations while they are in outdoor. If HIPs can successfully get sound information through some machine interface, dangerous situation will be avoided. Generally the profoundly deaf people do not use any hearing aid which does not provide any benefit. This paper presents, simple statistical features are used to classify the vehicle type and its distance based on sound signature recorded from the moving vehicles. An experimental protocol is designed to record the vehicle sound under different environment conditions and also at different speed of vehicles. Basic statistical features such as the standard deviation, Skewness, Kurtosis and frame energy have been used to extract the features. Probabilistic neural network (PNN) models are developed to classify the vehicle type and its distance. The effectiveness of the network is validated through stimulation.

Abstract: A probabilistic neural network (PNN) speech recognition model based on the partition clustering algorithm is proposed in this paper. The most important advantage of PNN is that training is easy and instantaneous. Therefore, PNN is capable of dealing with real time speech recognition. Besides, in order to increase the performance of PNN, the selection of data set is one of the most important issues. In this paper, using the partition clustering algorithm to select data is proposed. The proposed model is tested on two data sets from the field of spoken Arabic numbers, with promising results. The performance of the proposed model is compared to single back propagation neural network and integrated back propagation neural network. The final comparison result shows that the proposed model performs better than the other two neural networks, and has an accuracy rate of 92.41%.

Abstract: Epilepsy is one of the most common neurological disorders that greatly disturb patients’ daily lives. Traditional epileptic diagnosis relies on tedious visual screening by neurologists from lengthy EEG recording that requires the presence of seizure (ictal) activities. We proposed to study automated epileptic diagnosis using interictal EEG data that was much easier to collect than ictal data. The research aims to develop an automated diagnostic system that can use interictal EEG data to diagnose whether the person is epileptic. This system could also test epileptic seizures in order to provide doctors with further tests and potential monitor of patients. To test such a system, we extract power spectral feature, Petrosian fractal dimension, Higuchi fractal dimension and Hjorth parameters for analysis, from which we find our system can be used in patient monitoring(seizure detection) and seizure focus localization, with 98.333% and75.5% accuracy respectively.

Abstract: Nowadays, diagnosis for epilepsy depends on many systems helping the neurologists to quickly find interesting segments from the lengthy signal by automatic seizure detection. However, we notice that it is very difficult, to obtain long-term EEG data with seizure activities for epilepsy patients in areas lack of medical resources and trained neurologists. Therefore, we propose to study automated epileptic diagnosis using interictal EEG data that is much easier to collect than ictal data. The research, therefore, aims to develop an automated diagnostic system that can use interictal EEG data to diagnose whether the person is epileptic. To develop such a system, we extract from the EEG data three classes of features which respectively are Petrosian fractal dimension, Higuchi fractal dimension and Hjorth parameters and build a Probabilistic Neural Network (PNN) fed with these features. Meanwhile, we also broach demand for data standardization by analysis with EEG of epileptic patients.

Abstract: In view of poor Physical Properties,complex Pore Structure and high saturation of low porosity and low permeability gas layers ,in order to overcome the difficult of fluid property identification in low porosity and low permeability gas layers using conventional method, probabilistic neural network technique was Proposed.According to an example for low porosity and low permeability gas reservoil in Southwest China, Combined with well testing data,logging Response Characteristics of various fluid property layers were analyzed.According the correlation between logging Response Characteristic values and fluid property, PNN was trained and PNN prediction model was established. fluid property in the region layers were identified.The results showed that the PNN prediction model was very promising influid property identification.

Abstract: A new image feature selection method with the combination of Genetic Algorithm(GA) and Probabilistic Neural Network(PNN) is proposed and applied to potato shape feature selection and classification. The classifier selecting principle is investigated by combining with the genetic algorithm. A new feature selection method based on GA and PNN is put forward firstly. Comprehensively considering the factor of classification accuracy,selected feature number and the impact of the two factors, a new fitness function is proposed. The initial Zernike moments parameters of potatoes are optimized using improved genetic algorithm, and nineteen Zernike moments are extracted to form the shape feature. The shape detection accuracy can reach 93% and 100% respectively for the perfect and malformation potatoes. The tests indicate that the fitness function and feature selection method can be used for searching the best feature combination.

Abstract: Electronic Tongue is a kind of intelligent equipment which is used to distinguish tastes. An electronic tongue composed of a sensor array of ion-selective electrodes has been developed and used for the qualitative analysis of five different brands of mineral water. The acquired original data has been optimized by principal component analysis (PCA) and then the probabilistic neural network (PNN) model is designed to process the data. The application results show that the performance of the proposed method has surpasses the traditional BP neural network algorithm, the speed of recognition is fast and the accuracy rate can reach 100%, which gives the electronic tongue system good practicability and feasibility.

Abstract: To evaluate accurately working condition of guide, make maintenance strategy, and predict its residual life in the process of machining operation, a rolling guide rail condition monitoring system based on neural networks was constructed after key factors to guide life were investigated carefully. Eight B&K 4321 three-way vibration sensor were installed on slider surface to monitor the on-line condition of four guides and eight sliders. Vibration signals were processed by wavelet packet decomposition and the most sensitive features to guide life were selected by fuzzy clustering method. The relation between guide life and input vectors including vibration features and machining condition was built by radial basis probabilistic neural networks (RBPNN), which parameters were optimized by genetic algorithm. The experimental results show maximum forecast error is 360 hours and minimum forecast error is 63 hours.