[신간의 별자리x] 우리/미술, 그리고 ‘슬픔의 박물관’

Speech emotion recognition using deep 1D & 2D CNN LSTM networks

A survey on deep learning based face recognition

Information fusion is a key step in multimodal biometric systems. The fusion of information can occur at different levels of a recognition system, i.e., at the feature level, matching-score level, or decision level. However, feature level fusion is believed to be more effective owing to the fact that a feature set contains richer information about the input biometric data than the matching score or the output decision of a classifier. The goal of feature fusion for recognition is to combine relevant information from two or more feature vectors into a single one with more discriminative power than any of the input feature vectors. In pattern recognition problems, we are also interested in separating the classes. In this paper, we present discriminant correlation analysis (DCA), a feature level fusion technique that incorporates the class associations into the correlation analysis of the feature sets. DCA performs an effective feature fusion by maximizing the pairwise correlations across the two feature sets and, at the same time, eliminating the between-class correlations and restricting the correlations to be within the classes. Our proposed method can be used in pattern recognition applications for fusing the features extracted from multiple modalities or combining different feature vectors extracted from a single modality. It is noteworthy that DCA is the first technique that considers class structure in feature fusion. Moreover, it has a very low computational complexity and it can be employed in real-time applications. Multiple sets of experiments performed on various biometric databases and using different feature extraction techniques, show the effectiveness of our proposed method, which outperforms other state-of-the-art approaches.

/pdf/discriminant-correlation-analysis-real-time-feature-level-19198p50gn.pdf

Discriminant Correlation Analysis: Real-Time Feature Level Fusion for Multimodal Biometric Recognition

Plant diseases are important factors as they result in serious reduction in quality and quantity of agriculture products. Therefore, early detection and diagnosis of these diseases are important. To this end, we propose a deep learning-based approach that automates the process of classifying banana leaves diseases. In particular, we make use of the LeNet architecture as a convolutional neural network to classify image data sets. The preliminary results demonstrate the effectiveness of the proposed approach even under challenging conditions such as illumination, complex background, different resolution, size, pose, and orientation of real scene images.

A Deep Learning-based Approach for Banana Leaf Diseases Classification

This paper presents a method for automatic real time face and mouth recognition using radial basis function neural networks (RBFNN). The proposed method uses the motion information to localize the face region, and the face region is processed in YC"rC"b color space to determine the locations of the eyes. The center of the mouth is determined relative to the locations of the eyes. Facial and mouth features are extracted using multiscale morphological erosion and dilation operations, respectively. The facial features are extracted relative to the locations of the eyes, and mouth features are extracted relative to the locations of the eyes and mouth. The facial and mouth features are given as input to radial basis function neural networks. The RBFNN is used to recognize a person in video sequences using face and mouth modalities. The evidence from face and mouth modalities are combined using a weighting rule, and the result is used for identification and authentication. The performance of the system using facial and mouth features is evaluated in real time in the laboratory environment, and the system achieves a recognition rate (RR) of 99.0% and an equal error rate (EER) of about 0.73% for 50 subjects. The performance of the system is also evaluated for XM2VTS database, and the system achieves a recognition rate (RR) of 100% an equal error rate (EER) of about 0.25% for 50 subjects.

Real time face and mouth recognition using radial basis function neural networks

In agricultural field, paddy cultivation plays a vital role. But their growths are affected by various diseases. There will be decrease in the production, if the diseases are not identified at an early stage. The main goal of this work is to develop an image processing system that can identify and classify the various paddy plant diseases affecting the cultivation of paddy namely brown spot disease, leaf blast disease and bacterial blight disease. This work can be divided into two parts namely, paddy plant disease detection and recognition of paddy plant diseases. In disease detection, the disease affected portion of the paddy plant is first identified using Haar-like features and AdaBoost classifier. The detection accuracy rate is found to be 83.33%. In disease recognition, the paddy plant disease type is recognized using Scale Invariant Feature Transform (SIFT) feature and classifiers namely k-Nearest Neighbour (k-NN) and Support Vector Machine (SVM). By this approach one can detect the disease at an early stage and thus can take necessary steps in time to minimize the loss of production. The disease recognition accuracy rate is 91.10% using SVM and 93.33% using k-NN.

/pdf/detection-and-recognition-of-diseases-from-paddy-plant-leaf-1de8u2snfg.pdf

Detection and Recognition of Diseases from Paddy Plant Leaf Images

The paper presents a novel relative slope based algorithm for an on-line and off-line signature verification system capable of effectively establishing an individual's identify based solely on their handwriting characteristics. Current technologies in signature verification systems use various algorithms for feature point extraction, regression approach, Markov method, split and merge and genetic algorithms. We propose a slope based model, in which the input signature is divided into many segments using an optimized HMM method; then, the slope of every segment is calculated with respect to its previous segment, obtained after normalization of the signature. This feature of each segment is stored along with the two tier time metric information, which ensures lesser overhead with better performance while processing.

On-line and off-line signature verification using relative slope algorithm

At present situation, telecommunication department plays vital role in our day today human life. At the same time telecommunication area attains a rapid growth in very short period of years. Due to this fact, Telecom department is one of the senses deciding factor in world market. Because of this reason, this field turned out to be a most profitable area for investment. At the mean time competition between telecom industries is also there in the market. So churn prediction is very essential in telecom industries to retain their customers. This study focuses on data mining techniques for reducing customer churn and also likely to reduce error ratio.

http://ijsrp.org/research-paper-0414/ijsrp-p2848.pdf

Churn prediction in mobile telecom system using data mining techniques

This paper describes an efficient method to detect traffic lights and recognize their corresponding signal count-down timings. Driver Assistance System (DAS) are applications installed in vehicles that provide guidance for safe driving. Traffic Light detection is also a part of the DAS. The work is split into two parts of which, the first part deals with the detection of traffic lights. The changing traffic lights are detected by extracting the candidate region of traffic light and timer display by color segmentation using thresholding. The primary traffic light is detected by comparing the count of red pixels with the count of green pixels that is obtained as a result of thresholding. The second part recognizes the LED digits in the timer display by color segmentation, which is segmented and extracted by horizontal and vertical scanning. The extracted image is then resized to a standard size which is then converted into its binary form and bounding boxes of static width is placed over each segment of the extracted LED. The on/off status of each LED segment is obtained for all the seven segments and based on the different combination of each LED segment status, the digits can be recognized. DAS provides better realization of comfort, fuel economy and safety. The proposed method is able to detect the predominant traffic light and signal timings in a reasonable time. Experimental results show that an accuracy of 96% and 88% has been achieved for detection of traffic lights and recognition of LED digits respectively.

M. Balasubramanian

Papers

Real time face and mouth recognition using radial basis function neural networks

Detection and Recognition of Diseases from Paddy Plant Leaf Images

On-line and off-line signature verification using relative slope algorithm

Churn prediction in mobile telecom system using data mining techniques

Real time recognition of traffic light and their signal count-down timings