• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.11 no.12
• /
• pp.6038-6053
• /
• 2017

In this study, we propose an automatic melody extraction algorithm using deep learning. In this algorithm, feature images, generated using the energy of frequency band, are extracted from polyphonic audio files and a deep learning technique, a convolutional neural network (CNN), is applied on the feature images. In the training data, a short frame of polyphonic music is labeled as a musical note and a classifier based on CNN is learned in order to determine a pitch value of a short frame of audio signal. We want to build a novel structure of melody extraction, thus the proposed algorithm has a simple structure and instead of using various signal processing techniques for melody extraction, we use only a CNN to find a melody from a polyphonic audio. Despite of simple structure, the promising results are obtained in the experiments. Compared with state-of-the-art algorithms, the proposed algorithm did not give the best result, but comparable results were obtained and we believe they could be improved with the appropriate training data. In this paper, melody extraction and the proposed algorithm are introduced first, and the proposed algorithm is then further explained in detail. Finally, we present our experiment and the comparison of results follows.

• Journal of Korea Multimedia Society
• /
• v.24 no.11
• /
• pp.1552-1559
• /
• 2021

Convolution neural networks (CNNs) show notable performance in image processing and are used as representative core models. CNNs extract and learn features from large amounts of train dataset. In general, it has a structure in which a convolution layer and a fully connected layer are stacked. The core of CNN is the convolution layer. The size of the kernel used for feature extraction and the number that affect the depth of the feature map determine the amount of weight parameters of the CNN that can be learned. These parameters are the main causes of increasing the computational complexity and memory usage of the entire neural network. The most computationally expensive components in CNNs are fully connected and spatial convolution computations. In this paper, we propose a Fourier Convolution Neural Network that performs the operation of the convolution layer in the Fourier domain. We work on modifying and improving the amount of computation by applying the fast fourier transform method. Using the MNIST dataset, the performance was similar to that of the general CNN in terms of accuracy. In terms of operation speed, 7.2% faster operation speed was achieved. An average of 19% faster speed was achieved in experiments using 1024x1024 images and various sizes of kernels.

• International Journal of Computer Science & Network Security
• /
• v.23 no.8
• /
• pp.49-62
• /
• 2023

There has been a rapid increase in the creation and alteration of new malware samples which is a huge financial risk for many organizations. There is a huge demand for improvement in classification and detection mechanisms available today, as some of the old strategies like classification using mac learning algorithms were proved to be useful but cannot perform well in the scalable auto feature extraction scenario. To overcome this there must be a mechanism to automatically analyze malware based on the automatic feature extraction process. For this purpose, the dynamic analysis of real malware executable files has been done to extract useful features like API call sequence and opcode sequence. The use of different hashing techniques has been analyzed to further generate images and convert them into image representable form which will allow us to use more advanced classification approaches to classify huge amounts of images using deep learning approaches. The use of deep learning algorithms like convolutional neural networks enables the classification of malware by converting it into images. These images when fed into the CNN after being converted into the grayscale image will perform comparatively well in case of dynamic changes in malware code as image samples will be changed by few pixels when classified based on a greyscale image. In this work, we used VGG-16 architecture of CNN for experimentation.

• Korean Journal of Artificial Intelligence
• /
• v.11 no.4
• /
• pp.1-8
• /
• 2023

In this paper, we propose a black ice detection platform framework using Convolutional Neural Networks (CNNs). To overcome black ice problem, we introduce a real-time based early warning platform using CNN-based architecture, and furthermore, in order to enhance the accuracy of black ice detection, we apply a multi-scale dilation convolution feature fusion (MsDC-FF) technique. Then, we establish a specialized experimental platform by using a comprehensive dataset of thermal road black ice images for a training and evaluation purpose. Experimental results of a real-time black ice detection platform show the better performance of our proposed network model compared to conventional image segmentation models. Our proposed platform have achieved real-time segmentation of road black ice areas by deploying a road black ice area segmentation network on the edge device Jetson Nano devices. This approach in parallel using multi-scale dilated convolutions with different dilation rates had faster segmentation speeds due to its smaller model parameters. The proposed MsCD-FF Net(2) model had the fastest segmentation speed at 5.53 frame per second (FPS). Thereby encouraging safe driving for motorists and providing decision support for road surface management in the road traffic monitoring department.

• The Korean Journal of Food And Nutrition
• /
• v.37 no.4
• /
• pp.197-210
• /
• 2024

This study aims to develop a deep learning model to monitor rice serving amounts in institutional foodservice, enhancing personalized nutrition management. The goal is to identify the best convolutional neural network (CNN) for detecting rice quantities on serving trays, addressing balanced dietary intake challenges. Both a vanilla CNN and 12 pre-trained CNNs were tested, using features extracted from images of varying rice quantities on white trays. Configurations included optimizers, image generation, dropout, feature extraction, and fine-tuning, with top-1 validation accuracy as the evaluation metric. The vanilla CNN achieved 60% top-1 validation accuracy, while pre-trained CNNs significantly improved performance, reaching up to 90% accuracy. MobileNetV2, suitable for mobile devices, achieved a minimum 76% accuracy. These results suggest the model can effectively monitor rice servings, with potential for improvement through ongoing data collection and training. This development represents a significant advancement in personalized nutrition management, with high validation accuracy indicating its potential utility in dietary management. Continuous improvement based on expanding datasets promises enhanced precision and reliability, contributing to better health outcomes.

• Information Systems Review
• /
• v.22 no.4
• /
• pp.21-39
• /
• 2020

Recently, deep learning has shown high performance in various applications such as pattern analysis and image classification. Especially known as a difficult task in the field of machine learning research, stock market forecasting is an area where the effectiveness of deep learning techniques is being verified by many researchers. This study proposed a deep learning Convolutional Neural Network (CNN) model to predict the direction of stock prices. We then used the feature selection method to improve the performance of the model. We compared the performance of machine learning classifiers against CNN. The classifiers used in this study are as follows: Logistic Regression, Decision Tree, Neural Network, Support Vector Machine, Adaboost, Bagging, and Random Forest. The results of this study confirmed that the CNN showed higher performancecompared with other classifiers in the case of feature selection. The results show that the CNN model effectively predicted the stock price direction by analyzing the embedded values of the financial data

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.14 no.3
• /
• pp.1188-1203
• /
• 2020

The human expression contains a lot of information that can be used to detect complex conditions such as pain and fatigue. After deep learning became the mainstream method, the traditional feature extraction method no longer has advantages. However, in order to achieve higher accuracy, researchers continue to stack the number of layers of the neural network, which makes the real-time performance of the model weak. Therefore, this paper proposed an expression recognition framework based on densely concatenated convolutional neural networks to balance accuracy and latency and apply it to humanoid robots. The techniques of feature reuse and parameter compression in the framework improved the learning ability of the model and greatly reduced the parameters. Experiments showed that the proposed model can reduce tens of times the parameters at the expense of little accuracy.

• Journal of Broadcast Engineering
• /
• v.20 no.6
• /
• pp.938-949
• /
• 2015

In this paper, we propose a robust speed-limit sign recognition system which is durable to any sign changes caused by exterior damage or color contrast due to light direction. For recognition of speed-limit sign, we apply CNN which is showing an outstanding performance in pattern recognition field. However, original CNN uses multiple hidden layers to extract features and uses fully-connected method with MLP(Multi-layer perceptron) on the result. Therefore, the major demerit of conventional CNN is to require a long time for training and testing. In this paper, we apply randomly-connected classifier instead of fully-connected classifier by combining random forest with output of 2 layers of CNN. We prove that the recognition results of CNN with random forest show best performance than recognition results of CNN with SVM (Support Vector Machine) or MLP classifier when we use eight speed-limit signs of GTSRB (German Traffic Sign Recognition Benchmark).

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.24 no.3
• /
• pp.364-370
• /
• 2020

This paper is related to detection and classification for micro-sized aircraft that flies at low-altitude. The deep-learning based method using sounds coming from the micro-sized aircraft is proposed to detect and identify them efficiently. We use MFCC as sound features and CNN as a detector and classifier. We've proved that each micro-drones have their own distinguishable MFCC feature and confirmed that we can apply CNN as a detector and classifier even though drone sound has time-related sequence. Typically many papers deal with RNN for time-related features, but we prove that if the number of frame in the MFCC features are enough to contain the time-related information, we can classify those features with CNN. With this approach, we've achieved high detection and classification ratio with low-computation power at the same time using the data set which consists of four different drone sounds. So, this paper presents the simple and effecive method of detection and classification method for micro-sized aircraft.

• KIISE Transactions on Computing Practices
• /
• v.23 no.3
• /
• pp.194-198
• /
• 2017

In this paper, we propose a revised Deep Convolutional Neural Network (DCNN) model to extract Protein-Protein Interaction (PPIs) from the scientific literature. The proposed method has the merit of improving performance by applying various global features in addition to the simple lexical features used in conventional relation extraction approaches. In the experiments using AIMed, which is the most famous collection used for PPI extraction, the proposed model shows state-of-the art scores (78.0 F-score) revealing the best performance so far in this domain. Also, the paper shows that, without conducting feature engineering using complicated language processing, convolutional neural networks with embedding can achieve superior PPIE performance.