---

This paper introduces a novel approach to improve performance of speech recognition systems using a combination of features obtained from speech reconstructed phase space (RPS) and frequency domain analysis. By choosing an appropriate value for the dimension, reconstructed phase space is assured to be topologically equivalent to the dynamics of the speech production system, and could therefore include information that may be absent in analysis approaches based on linear methods. Also, complicated systems such as speech production system can present cyclic and oscillatory patterns and Poincare sections could be used as an effective tool in analysis of such trajectories. In this research, a statistical modeling approach based on Gaussian Mixture models (GMM) was applied to the Poincare sections of speech RPS. The final feature set is obtained from a feature selection stage omong parameters of GMM model and the usual Mel Frequency Cepstral coefficients (MFCC). An HMM-based speech recognition system and the TIMIT speech database are used to evaluate performance of the proposed feature extraction system for isolated and continuous speech recognition. Experiments represent about 5.7% absolute isolated phoneme recognition accuracy improvement in isolated phoneme recognition performance. The new approach is shown to be a viable and effective alternative to traditional feature extraction methods, particularly for signals such as speech with strong nonlinear characteristics.

---

In this paper a heuristic method, called Moving Window K-Nearest Neighbors (MW-KNN), for detecting QRS complexes was developed. To achieve this, a new simple 2-D geometrical feature space (feature space dimension was equal to 2) was extracted from the original electrocardiogram (ECG) signal. In this method, a sliding window was moved sample-by-sample on the preprocessed ECG signal. During each forward sliding, an artificial image was generated from the excerpted segment allocated in the window. Each image estimated by a 300×300 pixels matrix. Then, a pictorial-geometrical feature extraction technique based on curve-length was applied to each image for establishment of an appropriate feature space. Afterwards the K-Nearest Neighbors (KNN) Classification method was designed and implemented to the ECG signal. The proposed methods were applied to DAY general hospital high resolution holter data. For detection of QRS complex the average values of sensitivity Se = 99.93% and positive predictivity P+ = 99.88% were obtained.

---

The necessity to meet ongoing needs of industry, considering theoretical progress achievements and availability of cost-effective equipment, has encouraged numerous researchers to investigate the application of monitoring systems. In this paper the sound localization is implemented to find the impact position on the surface of a plate. As an experimental example the sound caused by ball impact on a ping pong table is used. For this purpose, a database is gathered. These sound's signals were recorded 25 times at 5 different points along the length of the table by a low cost microphone, attached to the surface. In the proposed method, first the data related to the ball impacts are detected and isolated from the whole pc recorded signals sent by the microphone. Then, the above 125 impacts are clustered based on the impact point locations, using a 4 dimensional space feature extracted from statistical signal moments. Furthermore in order to specify sound localization, a second space feature based on energy of wavelet transform coefficient signals was extracted. Ultimately for clustering the impact point locations, an artificial neural network was designed and applied to the above data. The results show average values of sensitivity Se=91.20% and positive predictivity P+=91.18%. Also, sensitivity Se=91.97% and positive predictivity P+‌=‌93.45%, correspondingly for impact localization.

---

Identification and classification of signals which are heard by underwater microphones (hydrophones) can be used extensively in harbor traffic management, especially in economical harbors. However, automatic identification and classification of acoustic signals which are received by passive sonar system is a challenging problem, because of variation in temporal and frequency characteristics of signals (even they are received from a same source). In this paper, a novel method for classification of acoustic signals is presented, based on DWT as preprocessing, a diverse range of feature extraction methods (principal component analysis and its variations (6 methods) and discriminant analysis and its variations (3 methods)), and 4 ensemble learning methods with 3 classifiers (multilayer perceptron (MLP), probabilistic neural network (PNN) and support vector machine (SVM)). Performing a diverse range of performance tests, the performances of different methods are assessed and the best ones are chosen for the proposed method. The proposed method is used to extract features and classify acoustic signals of 8 ships. Using the proposed method, some real signals and their noisy version are classified. The accuracy of the proposed method in classification of test signals with Gaussian white noise with -5, -10 and -15 signal-to-noise ratio is obtained as 99.83%, 97.06% and 83.56%, respectively.

---

In this study the sound localization is implemented to find the impact position on the surface of a glass plate using acoustical sensors. As an experimental example, the sound caused by ping pong ball impact on the glass plate is used. Most of the published paper algorithms are based on using large number of sensor with high sampling rates. In this study a new method is extended due to sound localization. In the proposed method, by reducing the number of sensors into two, a pattern for secondary points is extended. In the specified pattern, locations of points are restricted according to the sensors signal frequency specification. To achieve this goal, a database is gathered from sound caused by ball impact on the glass plate. Furthermore, in order to specify sound localization, space feature based on entropy of wavelet transform coefficient signals from frequency domain of impacts and geometrical specification was extracted. Finally by implementing signal processing into the data the location of impacts are specified. The results show average values of error and Standard deviation 17 centimeter and 1.34, respectively.

---

In this study, an application of support vector machine (SVM) for early fault detection in increasing the level of the start-up vessel in a Benson type once-through boiler during load changes is presented. The level increasing in the start-up vessel is happened due to thermal conditions disruption inside the boiler especially while the unit load is ramped-down. In this regard, first, the variables effective on increasing the level of start-up vessel was identified based on experimental data from a power plant unit, then the dimension of input variables was reduced by selecting appropriate features. Experimental results show that the hotwell surfaces’ temperature could be considered as the most appropriate indicator for steam quality deterioration. By comparing the extracted features from healthy and unhealthy conditions, appropriate fault model was developed using SVM with radial basis function (RBF) as the kernel. The performances of fault detection system were evaluated with respect to the similar faults at two different time periods happen in a steam power plant. The obtained results show the accuracy and feasibility of the proposed approach in early detection of faults during the unit’s load variations. Advantages of the proposed technique is preventing false alarm in power plants’ boilers as load changes.

---

Nowadays, diagnosis of diseases with high precision, high speed, low-cost and non-invasive approaches has become a necessity. In this regard, taking pulse signal is very easy and inexpensive, which due to the availability and feasibility of the process, can be very useful in the rapid diagnosis heart disease. If we can use the appropriate signal processing and intelligent methods in such a way that its accuracy and total cost equal those of other corresponding methods, we can say that we have reached a valuable achievement; in the current study we pursue the same purpose. In the first step, pressure pulse signals of 45 Coronary Arterial Disease (CAD) patients and 45 healthy persons are acquired from the left fingers using Task Force Monitor (TFM). Then the signals are filtered by wavelet transform (db6) and the wrong items are discarded. Then, the features corresponding to the CAD and healthy states are extracted which based on Time Domain Analysis. Finally, by choosing the best features, the data of healthy people and patients (CAD) are classified with Support Vector Machine (SVM) classifier by the accuracy rate of more than 85%.

---

In this paper, a new hybrid intelligent method is presented for detecting the bearing faults in the various rotating speeds. The vibration signals are collected in four conditions, including the normal state, the faulty inner race, the faulty outer race, and the faulty bearing element. Firstly, twenty-two statistical features in the time domain and four frequency features, three Wavelet packet decomposition (WPD), and the first five intrinsic mode functions obtained by the empirical mode decomposition (EMD) are extracted from the original signal; finally, the feature vector for each signal sample has 424 features. However, in the high dimensional feature matrix, there may exist the insensitive features to the presence of defects. Therefore, in this study, the compensation distance evaluation technique (CDET) is used to select the optimal features. Then, the selected features are used as the inputs of the support vector machine (SVM) classifier to diagnose the bearing conditions. In the CDET method, there is a threshold indicator that plays a decisive role in choosing the desired attributes. Also, the SVM method has some parameters that need to be set during the fault detection process. Therefore, the particle swarm optimization (PSO) algorithm is used to determine the optimal threshold in the CDET method and the optimal SVM parameters, so that the prediction error of the bearing conditions and the number of the selected features are minimized. The obtained results demonstrate that the selected features are well able to differentiate between different bearing conditions at various speeds. Comparing the results of this paper with other fault detection methods indicates the ability of the proposed method.

---

   Civil structures may experience unexpected loads and consequently damages during their life cycle. Damage identification has been a challenging inverse problem in structural health monitoring. The main difficulty is characterizing the unknown relation between the measurements and damage patterns. Such damage indicators would ideally be able to identify the existence, location, and severity of damages. In order to solve such problems, biologically inspired soft-computing techniques have gained traction. The most widely used soft-computing method, called neural networks is designed such that it can learn from data without a need of feature design process. Damage pattern can be detected using neural network. A deep unsupervised neural network can recognize patterns and extract features from data. In this paper a methodology is described for global and local health condition assessment of structural systems using vibration response of the structure. The model incorporates Fast Fourier Transform and unsupervised deep Boltzmann machine to extract features from the frequency domain of the recorded signals. Restricted boltzmann machine is a shallow neural network with two layer. First layer of restricted boltzmann machine called input layer and second layer of restricted boltzmann machine called hidden layer.Deep Boltzmann machine created by setting some restricted Boltzmann machine sequentional. Hidden layer of each restricted boltzmann machine is input layer of next restricted boltzmann machine. Each layer of restricted Boltzmann machine extract features form input data Recorded data divided to smaller vectors. Fast fourier transformation used to transform divided vectors into frequency domain.  A benefit of the proposed model is that it does not require costly experimental results to be obtained from a scaled version of the structure to simulate different damage states of the structure and only vibration response of the healthy structure is needed to training deep neural network. The input consists of a set of records obtained from the healthy state of the structure and another set of records with unknown health states. The model extracts information from both healthy and unknown sets to determine the health states of the unknown set. The healthy records are low intensity vibrations of the structure at least in one planar direction in the healthy state in the form of time series signals and The unknown records are low intensity vibrations of the structure on unknown state of health. Ambient vibrations can be due to wind, traffic, or human/pedestrian activities. An appropiate health index is defined and calculated for each part of the structure. The value of this index is between 0 and 1. The closer the value is to 1 the healthier the structure. To evaluate the efficiency of the proposed method a building structures with 35 story has been simulated in OPENSEES. Data collection should be selected appropriately to prevent errors. Obtained result demonstrate that proposed method has about 95 percent efficiency to predict damages and their severity. Different damage state put on due to three earthquakes with different severity. Structural health index calculated after each earthquake. Calculated structural health index demonstrate efficieency of proposed method for detecting damages and severity of damages.

---

Fualts in rolling bearings are one of the main reasons for the failure of rotating machinery. faults detection rolling bearing has played an essential role in the reliable performance of production units. In addition, condition monitoring of machinery using vibration analysis is one of the most powerful tools in measuring the health of mechanical systems. This research proposes an intelligent system for detecting defects in rolling bearings based on vibration analysis. In the intelligent faults detection system, the extracted features of the vibration signals in the time domain and the radial basis function neural network are used. The train and test datasets are presented to the radial basis function neural network intelligent system. The results of neural network learning show the very successful performance of the intelligent fault diagnosis system in detecting the health state and triple fault states of rolling bearings