• Home
  • Hassan Farsi
    • List of Articles Hassan Farsi

      • Open Access Article

        1 - Fast Automatic Face Recognition from Single Image per Person Using GAW-KNN
        Hassan Farsi Mohammad Hasheminejad
        Real time face recognition systems have several limitations such as collecting features. One training sample per target means less feature extraction techniques are available to use. To obtain an acceptable accuracy, most of face recognition algorithms need more than on Full Text
        Real time face recognition systems have several limitations such as collecting features. One training sample per target means less feature extraction techniques are available to use. To obtain an acceptable accuracy, most of face recognition algorithms need more than one training sample per target. In these applications, accuracy of recognition dramatically reduces for the case of one training sample per target face image because of head rotation and variation in illumination state. In this paper, a new hybrid face recognition method by using single image per person is proposed, which is robust against illumination variations. To achieve robustness against head variations, a rotation detection and compensation stage is added. This method is called Weighted Graphs and PCA (WGPCA). It uses harmony of face components to extract and normalize features, and genetic algorithm with a training set is used to learn the most useful features and real-valued weights associated to individual attributes in the features. The k-nearest neighbor algorithm is applied to classify new faces based on their weighted features from the templates of the training set. Each template contains the corrected distances (Graphs) of different points on the face components and the results of Principal Component Analysis (PCA) applied to the output of face detection rectangle. The proposed hybrid algorithm is trained using MATLAB software to determine best features and their associated weights and is then implemented by using delphi XE2 programming environment to recognize faces in real time. The main advantage of this algorithm is the capability of recognizing the face by only one picture in real time. The obtained results of the proposed technique on FERET database show that the accuracy and effectiveness of the proposed algorithm. Manuscript Document
      • Open Access Article

        2 - Video Transmission Using New Adaptive Modulation and Coding Scheme in OFDM based Cognitive Radio
        Hassan Farsi Farid Jafarian
        As Cognitive Radio (CR) used in video applications, user-comprehended video quality practiced by secondary users is an important metric to judge effectiveness of CR technologies. We propose a new adaptive modulation and coding (AMC) scheme for CR, which is OFDM based sy Full Text
        As Cognitive Radio (CR) used in video applications, user-comprehended video quality practiced by secondary users is an important metric to judge effectiveness of CR technologies. We propose a new adaptive modulation and coding (AMC) scheme for CR, which is OFDM based system that is compliant with the IEEE.802.16. The proposed CR alters its modulation and coding rate to provide high quality system. In this scheme, CR using its ability to consciousness of various parameters including knowledge of the white holes in the channel spectrum via channel sensing, SNR, carrier to interference and noise ratio (CINR), and Modulation order Product code Rate (MPR) selects an optimum modulation and coding rate. In this scheme, we model the AMC function using Artificial Neural Network (ANN). Since AMC is naturally a non-liner function, ANN is selected to model this function. In order to achieve more accurate model, Genetic algorithm (GA) and Particle Swarm Optimization (PSO) are selected to optimize the function representing relationship between inputs and outputs of ANN, i.e., AMC model. Inputs of ANN are CR knowledge parameters, and the outputs are modulation type and coding rate. Presenting a perfect AMC model is advantage of this scheme because of considering all impressive parameters including CINR, available bandwidth, SNR and MPR to select optimum modulation and coding rate. Also, we show that in this application, GA rather than PSO is better choice for optimization algorithm. Manuscript Document
      • Open Access Article

        3 - Image Retrieval Using Color-Texture Features Extracted From Gabor-Walsh Wavelet Pyramid
        Sajad Mohammadzadeh Hassan Farsi
        Image retrieval is one of the most applicable image processing techniques which have been extensively used. Feature extraction is one of the most important procedures used for interpretation and indexing images in Content-Based Image Retrieval (CBIR) systems. Effective Full Text
        Image retrieval is one of the most applicable image processing techniques which have been extensively used. Feature extraction is one of the most important procedures used for interpretation and indexing images in Content-Based Image Retrieval (CBIR) systems. Effective storage, indexing and managing a large number of image collections are critical challenges in computer systems. There are many proposed methods to overcome these problems. However, the rate of image retrieval and speed of retrieval are still interesting fields of researches. In this paper, we propose a new method based on combination of Gabor filter and Walsh transform and Wavelet Pyramid (GWWP). The Crossover Point (CP) of precision and recall are considered as metrics to evaluate and compare different methods. The Obtained results show using GWWP provides better performance in compared to with other methods. Manuscript Document
      • Open Access Article

        4 - Node to Node Watermarking in Wireless Sensor Networks for Authentication of Self Nodes
        Hassan Farsi Seyed Morteza Nourian
        In order to solve some security issues in Wireless Sensor Networks (WSNs), node to node authentication method based on digital watermarking technique for verification of relative nodes is proposed. In the proposed method, some algorithms with low computational for gener Full Text
        In order to solve some security issues in Wireless Sensor Networks (WSNs), node to node authentication method based on digital watermarking technique for verification of relative nodes is proposed. In the proposed method, some algorithms with low computational for generation, embedding and detection of security ID are designed. The collected data packets by the nodes are marked using security ID. In the proposed method, header is used to mark the packets. Since the nature of the sensor networks is cooperative, using the head of the packets is proposed for authentication. Also using the marked head can prevent from sending and receiving fake data in the other nodes. Simulations have been performed in environments with imposing unrealistic data and having a probability from 1% to 10%. Comparing the proposed method with other methods shows that the proposed method in term of security, reducing traffic and increasing network lifetime is more effective. Manuscript Document
      • Open Access Article

        5 - Fusion Infrared and Visible Images Using Optimal Weights
        Mehrnoush  Gholampour Hassan Farsi Sajad Mohammadzadeh
        Image fusion is a process in which different images recorded by several sensors from one scene are combined to provide a final image with higher quality compared to each individual input image. In fact, combination of different images recorded by different sensors is on Full Text
        Image fusion is a process in which different images recorded by several sensors from one scene are combined to provide a final image with higher quality compared to each individual input image. In fact, combination of different images recorded by different sensors is one of image fusion methods. The fusion is performed based on maintaining useful features and reducing or removing useless features. The aim of fusion has to be clearly specified. In this paper we propose a new method which combines vision and infrared images by weighting average to provide better image quality. The weighting average is performed in gradient domain. The weight of each image depends on its useful features. Since these images are recorded in night vision, the useful features are related to clear scene details. For this reason, object detection is applied on the infrared image and considered as its weight. The vision image is also considered as a complementary of infrared image weight. The averaging is performed in gradient of input images, and final composed image is obtained by Gauss-Seidel method. The quality of resulted image by the proposed algorithm is compared to the obtained images by state-of-the-art algorithms using quantitative and qualitative measures. The obtained results show that the proposed algorithm provides better image quality. Manuscript Document
      • Open Access Article

        6 - Instance Based Sparse Classifier Fusion for Speaker Verification
        Mohammad Hasheminejad Hassan Farsi
        This paper focuses on the problem of ensemble classification for text-independent speaker verification. Ensemble classification is an efficient method to improve the performance of the classification system. This method gains the advantage of a set of expert classifiers Full Text
        This paper focuses on the problem of ensemble classification for text-independent speaker verification. Ensemble classification is an efficient method to improve the performance of the classification system. This method gains the advantage of a set of expert classifiers. A speaker verification system gets an input utterance and an identity claim, then verifies the claim in terms of a matching score. This score determines the resemblance of the input utterance and pre-enrolled target speakers. Since there is a variety of information in a speech signal, state-of-the-art speaker verification systems use a set of complementary classifiers to provide a reliable decision about the verification. Such a system receives some scores as input and takes a binary decision: accept or reject the claimed identity. Most of the recent studies on the classifier fusion for speaker verification used a weighted linear combination of the base classifiers. The corresponding weights are estimated using logistic regression. Additional researches have been performed on ensemble classification by adding different regularization terms to the logistic regression formulae. However, there are missing points in this type of ensemble classification, which are the correlation of the base classifiers and the superiority of some base classifiers for each test instance. We address both problems, by an instance based classifier ensemble selection and weight determination method. Our extensive studies on NIST 2004 speaker recognition evaluation (SRE) corpus in terms of EER, minDCF and minCLLR show the effectiveness of the proposed method. Manuscript Document
      • Open Access Article

        7 - Improved Generic Object Retrieval In Large Scale Databases By SURF Descriptor
        Hassan Farsi Reza Nasiripour Sajad Mohammadzadeh
        Normally, the-state-of-the-art methods in field of object retrieval for large databases are achieved by training process. We propose a novel large-scale generic object retrieval which only uses a single query image and training-free. Current object retrieval methods req Full Text
        Normally, the-state-of-the-art methods in field of object retrieval for large databases are achieved by training process. We propose a novel large-scale generic object retrieval which only uses a single query image and training-free. Current object retrieval methods require a part of image database for training to construct the classifier. This training can be supervised or unsupervised and semi-supervised. In the proposed method, the query image can be a typical real image of the object. The object is constructed based on Speeded Up Robust Features (SURF) points acquired from the image. Information of relative positions, scale and orientation between SURF points are calculated and constructed into the object model. Dynamic programming is used to try all possible combinations of SURF points for query and datasets images. The ability to match partial affine transformed object images comes from the robustness of SURF points and the flexibility of the model. Occlusion is handled by specifying the probability of a missing SURF point in the model. Experimental results show that this matching technique is robust under partial occlusion and rotation. The properties and performance of the proposed method are demonstrated on the large databases. The obtained results illustrate that the proposed method improves the efficiency, speeds up recovery and reduces the storage space. Manuscript Document
      • Open Access Article

        8 - Eye Gaze Detection Based on Learning Automata by Using SURF Descriptor
        Hassan Farsi Reza Nasiripour Sajad Mohammadzadeh
        In the last decade, eye gaze detection system is one of the most important areas in image processing and computer vision. The performance of eye gaze detection system depends on iris detection and recognition (IR). Iris recognition is very important role for person iden Full Text
        In the last decade, eye gaze detection system is one of the most important areas in image processing and computer vision. The performance of eye gaze detection system depends on iris detection and recognition (IR). Iris recognition is very important role for person identification. The aim of this paper is to achieve higher recognition rate compared to learning automata based methods. Usually, iris retrieval based systems consist of several parts as follows: pre-processing, iris detection, normalization, feature extraction and classification which are captured from eye region. In this paper, a new method without normalization step is proposed. Meanwhile, Speeded up Robust Features (SURF) descriptor is used to extract features of iris images. The descriptor of each iris image creates a vector with 64 dimensions. For classification step, learning automata classifier is applied. The proposed method is tested on three known iris databases; UBIRIS, MMU and UPOL database. The proposed method results in recognition rate of 100% for UBIRIS and UPOL databases and 99.86% for MMU iris database. Also, EER rate of the proposed method for UBIRIS, UPOL and MMU iris database are 0.00%, 0.00% and 0.008%, respectively. Experimental results show that the proposed learning automata classifier results in minimum classification error, and improves precision and computation time. Manuscript Document
      • Open Access Article

        9 - Improvement in Accuracy and Speed of Image Semantic Segmentation via Convolution Neural Network Encoder-Decoder
        هانیه زمانیان Hassan Farsi Sajad Mohammadzadeh
        Recent researches on pixel-wise semantic segmentation use deep neural networks to improve accuracy and speed of these networks in order to increase the efficiency in practical applications such as automatic driving. These approaches have used deep architecture to predic Full Text
        Recent researches on pixel-wise semantic segmentation use deep neural networks to improve accuracy and speed of these networks in order to increase the efficiency in practical applications such as automatic driving. These approaches have used deep architecture to predict pixel tags, but the obtained results seem to be undesirable. The reason for these unacceptable results is mainly due to the existence of max pooling operators, which reduces the resolution of the feature maps. In this paper, we present a convolutional neural network composed of encoder-decoder segments based on successful SegNet network. The encoder section has a depth of 2, which in the first part has 5 convolutional layers, in which each layer has 64 filters with dimensions of 3×3. In the decoding section, the dimensions of the decoding filters are adjusted according to the convolutions used at each step of the encoding. So, at each step, 64 filters with the size of 3×3 are used for coding where the weights of these filters are adjusted by network training and adapted to the educational data. Due to having the low depth of 2, and the low number of parameters in proposed network, the speed and the accuracy improve compared to the popular networks such as SegNet and DeepLab. For the CamVid dataset, after a total of 60,000 iterations, we obtain the 91% for global accuracy, which indicates improvements in the efficiency of proposed method. Manuscript Document
      • Open Access Article

        10 - Social Groups Detection in Crowd by Using Automatic Fuzzy Clustering with PSO
        Ali Akbari Hassan Farsi Sajad Mohammadzadeh
        Detecting social groups is one of the most important and complex problems which has been concerned recently. This process and relation between members in the groups are necessary for human-like robots shortly. Moving in a group means to be a subsystem in the group. In o Full Text
        Detecting social groups is one of the most important and complex problems which has been concerned recently. This process and relation between members in the groups are necessary for human-like robots shortly. Moving in a group means to be a subsystem in the group. In other words, a group containing two or more persons can be considered to be in the same direction of movement with the same speed of movement. All datasets contain some information about trajectories and labels of the members. The aim is to detect social groups containing two or more persons or detecting the individual motion of a person. For detecting social groups in the proposed method, automatic fuzzy clustering with Particle Swarm Optimization (PSO) is used. The automatic fuzzy clustering with the PSO introduced in the proposed method does not need to know the number of groups. At first, the locations of all people in frequent frames are detected and the average of locations is given to automatic fuzzy clustering with the PSO. The proposed method provides reliable results in valid datasets. The proposed method is compared with a method that provides better results while needs training data for the training step, but the proposed method does not require training at all. This characteristic of the proposed method increases the ability of its implementation for robots. The indexing results show that the proposed method can automatically find social groups without accessing the number of groups and requiring training data at all. Manuscript Document