Search

Although CNN shows high performance in the image recognition field, it has great disadvantages. They are to take a long time to perform machine learning due to a lack of system resources, and to consume a lot of power due to the great volume of computation. A convolution layer is a key element of convolutional neural network processing. This paper presents a convolution layer hardware using processing elements with a low-power multiplier. The low-power multiplier reduces the switching activity by increasing the exchange rate between the multiplier and multiplicand. The proposed convolution layer hardware was implemented on the Intel DE1-SoC FPGA Board using Verilog-HDL. The performance was verified by comparing the exchange rate with the existing multipliers when performing multiplications to process convolutional operations for the MNIST image database. [ABSTRACT FROM AUTHOR]

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. [ABSTRACT FROM AUTHOR]

The convolutional neural network algorithm (CNN) using MNIST's handwritten digit dataset has very good classification performance for digits that are tilted to the level of tilt of the learned digits, but its classification performance for digits that are tilted significantly more than the level drops sharply. This paper identifies the lower classification performance of the CNN depending on the tilt of digit and proposes an improved algorithm using indirect tilt correction for them. The proposed algorithm generates tilt-corrected digit candidates of an input digit and creates their CNN output map. It then designates a class of the output map that matches the CNN output characteristics of the upwritten digits as a class of the input digit. Compared to the CNN, the proposed algorithm decreased classification accuracy by about 1% for digits tilted less than 10o, but increased classification accuracy by about 10% to 69% for digits tilted more than 30o . [ABSTRACT FROM AUTHOR]

This paper proposes a CNN accelerator which is optimized Pooling layer operation incorporated in Multiplication And Accumulation(MAC) to reduce the memory size. For optimizing memory and data path circuit, the quantized 8bit integer weights are used instead of 32bit floating-point weights for pre-training of MNIST data set. To reduce chip area, the proposed CNN model is reduced by a convolutional layer, a 4*4 Max Pooling, and two fully connected layers. And all the operations use specific MAC with approximation adders and multipliers. 94% of internal memory size reduction is achieved by simultaneously performing the convolution and the pooling operation in the proposed architecture. The proposed accelerator chip is designed by using TSMC65nmGP CMOS process. That has about half size of our previous paper, 0.8*0.9 = 0.72mm² . The presented CNN accelerator chip achieves 94% accuracy and 77us inference time per an MNIST image. [ABSTRACT FROM AUTHOR]

Arrhythmia is a condition in which the heart has an irregular rhythm or abnormal heart rate, early diagnosis and management is very important because it can cause stroke, cardiac arrest, or even death. In this paper, we propose arrhythmia classification using hybrid combination model of CNN-BLSTM. For this purpose, the QRS features are detected from noise removed signal through pre-processing and a single bit segment was extracted. In this case, the GAN oversampling technique is applied to solve the data imbalance problem. It consisted of CNN layers to extract the patterns of the arrhythmia precisely, used them as the input of the BLSTM. The weights were learned through deep learning and the learning model was evaluated by the validation data. To evaluate the performance of the proposed method, classification accuracy, precision, recall, and F1-score were compared by using the MIT-BIH arrhythmia database. The achieved scores indicate 99.30%, 98.70%, 97.50%, 98.06% in terms of the accuracy, precision, recall, F1 score, respectively. [ABSTRACT FROM AUTHOR]

Music markets have entered the era of streaming. In order to select and propose music that suits the taste of music consumers, there is an active demand and research on an automatic music genre classification system. We propose a method to improve the accuracy of genre unclassified songs, which was a lack of the previous system, by using a generative adversarial network (GAN) to further develop the automatic voting system for deep learning music genre using Softmax proposed in the previous paper. In the previous study, if the spectrogram of the song was ambiguous to grasp the genre of the song, it was forced to leave it as an unclassified song. In this paper, we proposed a system that increases the accuracy of genre classification of unclassified songs by converting the spectrogram of unclassified songs into an easy-to-read spectrogram using GAN. And the result of the experiment was able to derive an excellent result compared to the existing method. [ABSTRACT FROM AUTHOR]

The ability to extract useful information from an image, such as the human eye, is an interface technology essential for AI computer implementation. The building recognition technology has a lower recognition rate than other image recognition technologies due to the various building shapes, the ambient noise images according to the season, and the distortion by angle and distance. The computer vision based building recognition algorithms presented so far has limitations in discernment and expandability due to manual definition of building characteristics. This paper introduces the deep learning CNN (Convolutional Neural Network) model, and proposes new method to improve the recognition rate even by changes of building images caused by season, illumination, angle and perspective. This paper introduces the partial images that characterize the building, such as windows or wall images, and executes the training with whole building images. Experimental results show that the building recognition rate is improved by about 14% compared to the general CNN model. [ABSTRACT FROM AUTHOR]

A convolutional neural network(CNN) is widely used in the computer vision tasks, but its computing power requirement needs a design of a special circuit. Most of the computations in a CNN can be implemented efficiently in a digital circuit, but the SoftMax layer has operations unsuitable for circuit implementation, which are exponential and logarithmic functions. This paper proposes a new method to integrate the exponential and logarithmic tables of the conventional circuits into a single table. The proposed structure accesses a look-up table (LUT) only with a few maximum values, and the LUT has the result value directly. Our proposed method significantly reduces the space complexity of the SoftMax layer circuit implementation. But our resulting circuit is comparable to the original baseline with small degradation in precision. [ABSTRACT FROM AUTHOR]

In this paper, we propose a classification method of pulsed radar signal attributes by applying machine learning to a convolutional neural network and a long short-term memory(CNN-LSTM) network when the radio frequency and pulse repetition interval of the signal changes with a long period. CNN has a function to extract data features, and LSTM shows good classification performance when sequential data are highly correlated. However, LSTM has a problem that performance is degraded when the length of the input data is long. In the proposed method, CNN is applied to extract compressed features from long-period data to reduce the length of inputs to LSTM. In simulation results, the proposed method shows more than 95% correct classification rate when the data drop rate is 10% for 182,516 total radar signal attribute combinations. [ABSTRACT FROM AUTHOR]

The limitations of applying a variety of artificial intelligence to the medical community are, first, subjective views, extensive interpreters and physical fatigue in interpreting the image of an interpreter's illness. And there are questions about how long it takes to collect annotated data sets for each illness and whether to get sufficient training data without compromising the performance of the developed deep learning algorithm. In this paper, when collecting basic images based on acne data sets, the selection criteria and collection procedures are described, and a model is proposed to classify data into small loss rates (5.46%) and high accuracy (96.26%) in the sequential structure. The performance of the proposed model is compared and verified through a comparative experiment with the model provided by Keras. Similar phenomena are expected to be applied to the field of medical and skin care by applying them to the acne classification model proposed in this paper in the future. [ABSTRACT FROM AUTHOR]

Flood damage can cause floods or tsunamis, which can result in enormous loss of life and property. In this regard, damage can be reduced by making a quick evacuation decision through flood prediction, and many studies are underway in this field to predict floods using time series data. In this paper, we propose a CNN-based time series prediction model. A CNN-based water level prediction model was implemented using the river level and precipitation, and the performance was confirmed by comparing it with the LSTM and GRU models, which are often used for time series prediction. In addition, by checking the performance difference according to the size of the input data, it was possible to find the points to be supplemented, and it was confirmed that better performance than LSTM and GRU could be obtained. Through this, it is thought that it can be utilized as an initial study for flood prediction. [ABSTRACT FROM AUTHOR]

Individual mental fatigue not only reduces cognitive ability and work performance, but also becomes a major factor in large and small accidents occurring in daily life. In this paper, a CNN model for EEG-based mental fatigue discrimination was proposed. To this end, EEG in the resting state and task state were collected and applied to the proposed CNN model, and then the model performance was analyzed. All subjects who participated in the experiment were right-handed male students attending university, with and average age of 25.5 years. Spectral analysis was performed on the measured EEG in each state, and the performance of the CNN model was compared and analyzed using the raw EEG, absolute power, and relative power as input data of the CNN model. As a result, the relative power of the occipital lobe position in the alpha band showed the best performance. The model accuracy is 85.6% for training data, 78.5% for validation, and 95.7% for test data. The proposed model can be applied to the development of an automated system for mental fatigue detection. [ABSTRACT FROM AUTHOR]

Recent advance in low cost and light-weight drones has extended their application areas in both military and private sectors. Accordingly surveillance program against unfriendly drones has become an important issue. Drone detection and classification technique has long been emphasized in order to prevent attacks or accidents by commercial drones in urban areas. Most commercial drones have small sizes and low reflection and hence typical sensors that use acoustic, infrared, or radar signals exhibit limited performances. Recently, artificial intelligence algorithm has been actively exploited to enhance radar image identification performance. In this paper, we adopt machined learning algorithm for high resolution radar imaging in drone detection and classification applications. For this purpose, simulation is carried out against commercial drone models and compared with experimental data obtained through high resolution radar field test. [ABSTRACT FROM AUTHOR]

This paper presents an automatic modulation classification method that involves the application of various imaging algorithms to a convolutional neural network (CNN). The effect of time-series data imaging on the performance of CNN-based modulation classification is analyzed. Our experiment suggests that converting raw signal data into image data using Markov transition field can reduce the error rate of CNN classification from 34 % to 30 % in case of −6 dB signal to noise ratio (SNR) and from 37 % to 18 % in case of 0 dB SNR. This study shows that time-series imaging is a viable preprocessing method for improving the performance of CNN-based modulation classification. [ABSTRACT FROM AUTHOR]

In a CNN for image classification, a SoftMax layer is usually placed at the end. The exponentinal and logarithmic operations in the SoftMax layer are not adequate to be implemented in an accelerator circuit. The operations are usually implemented with look-up tables, and the exponential operation can be implemented in an iterative method. This paper proposes a successive approximation method to calculate a logarithm to remove a very large look-up table. By substituing the large table with two very small tables, the circuit can be reduced much. The experimental results show that the 85% area reduction can be reached with a small error degradation. [ABSTRACT FROM AUTHOR]

This paper presents a machine learning (ML) algorithm to detect orthogonal frequency-division multiplexing (OFDM) signals. Spectrum sensing is a key technology in cognitive radio communication, which enhances spectral efficiency. Recently, signals transmitted and received by wireless communication systems have been based on OFDM. These signals contain a pilot signal for channel calibration, which can be detected using a spectral correlation function (SCF). In this process, the FAM algorithm is applied for efficient SCF operation herein. The existence of OFDM signals is determined through a convolutional neural network-based ML algorithm using the SCF as input data. The learning data for ML use the SCF values for OFDM signals with signal-to-noise ratios (SNRs) of −20 to 0 dB. Consequently, on evaluation of the post-learning optimized neural network performance, signals were detected with a probability of 0.9 at the condition of 0.08 false alarm probability for reception signals with SNRs of −12 dB. [ABSTRACT FROM AUTHOR]

The technology of interiors is currently developing around the world. According to various studies, the use of plants to create an environment in the home interior is increasing. However, households using furniture are designed as environment-friendly environment interiors, and in Korea and abroad, plants are used for home interiors. Unexpected accidents are occurring. As a result, there were books and broadcasts about the dangers of specific plants, but until now, accidents continue to occur because they do not properly recognize the dangers of specific plants. Therefore, in this paper, we propose a toxic plant identification system based on a multiplicative neural network model that identifies common toxic plants commonly found in Korea. We propose a high efficiency model. Through this, toxic plants can be identified with higher accuracy and safety accidents caused by toxic plants. [ABSTRACT FROM AUTHOR]

Recently, due to the proliferation of IoT sensors, the development of big data and artificial intelligence, time series prediction research on fine dust pollution is actively conducted. However, because the data representing fine dust contamination changes rapidly, traditional time series prediction methods do not provide a level of accuracy that can be used in the field. In this paper, we propose a method that reflects the classification results of environmental conditions through CNN when predicting micro dust contamination using LSTM. Although LSTM and CNN are independent, they are integrated into one network through the interface, so this method is easier to understand than the application LSTM. In the verification experiments of the proposed method using Beijing PM2.5 data, the prediction accuracy and predictive power for the timing of change were consistently improved in various experimental cases. [ABSTRACT FROM AUTHOR]

This paper proposes an image prediction system architecture for deep running in enterprise environment. Easily transform into an artificial intelligence platform for an enterprise environment, and allow sufficient deep-running services to be developed and modified even in Java-centric architectures to improve the shortcomings of Java-centric enterprise development because artificial intelligence platforms are concentrated in the pipeline. In addition, based on the proposed environment, we propose a more accurate prediction system in the deep running architecture environment that has been previously learned through image forecasting experiments. Experiments show 95.23% accuracy in the image example provided for deep running to be performed, and the proposed model shows 96.54% accuracy compared to other similar models. [ABSTRACT FROM AUTHOR]

Convolutional neural networks (CNNs) show high performance in the computer vision, but they require an enormous amount of operations, making them unsuitable for some resource- or energy-starving environments like the embedded environments. To overcome this problem, there have been much research on accelerators or pruning of CNNs. The previous pruning schemes have not considered the architecture of CNN accelerators, so the accelerators for the pruned CNNs have some inefficiency. This paper proposes a new pruning scheme, Conv-XP, which considers the architecture of CNN accelerators. In Conv-XP, the pruning is performed following the ‘X’ or ‘+’ shape. The Conv-XP scheme induces a simple architecture of the CNN accelerators. The experimental results show that the Conv-XP scheme does not degrade the accuracy of CNNs, and that the accelerator area can be reduced by 12.8%. [ABSTRACT FROM AUTHOR]

The current TW3 - based maximum height prediction technique used in KMAA(Korean Medical Academy of Auxology) is manual and subjective, and it requires a lot of time and effort in the medical treatment, while the interest in the child 's growth is very high. In addition, the technique of classifying images using deep learning, especially convolutional neural networks, is used in many 瓦elds at a more accurate level than the human eyes, also there is no exception in the medical 瓦eld. In this paper, we introduce a TW3 algorithm using deep learning, that uses the convolutional neural network to predict the growth level of the left hand bone, to predict the maximum height of child and youth in order to increase the reliability of predictions and improve the convenience of the doctor. [ABSTRACT FROM AUTHOR]

Online defamation incidents such as Internet news comments on portal sites, SNS, and community sites are increasing in recent years. Bias and hate expressions threaten online service users in various forms, such as invasion of privacy and personal attacks, and defamation issues. In the past few years, academia and industry have been approaching in various ways to solve this problem The purpose of this study is to build a dataset and experiment with deep learning classification modeling for detecting various bias expressions as well as hate expressions. The dataset was annotated 7 labels that 10 personnel cross-checked. In this study, each of the 7 classes in a dataset of about 137,111 Korean internet news comments is binary classified and analyzed through deep learning techniques. The Proposed technique used in this study is multi-channel CNN model with attention. As a result of the experiment, the weighted average f1 score was 70.32% of performance. [ABSTRACT FROM AUTHOR]

The sound collected through the multi-rotor unmanned aerial vehicle (UAV) includes the ego noise generated by the motor or propeller, or the wind noise generated during the flight, and thus the quality is greatly impaired. In a multi-rotor UAV environment, both the magnitude and phase of the target sound are greatly corrupted, so it is necessary to enhance the sound in consideration of both the magnitude and phase. However, it is difficult to improve the phase because it does not show the structural characteristics. in this study, we propose a CNN-based complex spectrogram enhancement method that removes noise based on complex spectrogram that can represent both magnitude and phase. Experimental results reveal that the proposed method improves enhancement performance by considering both the magnitude and phase of the complex spectrogram. [ABSTRACT FROM AUTHOR]

Research that implements the classification process through Deep Learning algorithm, one of the outstanding human abilities, includes a unimodal model, a multi-modal model, and a multi-modal method using music videos. In this study, the results were better by suggesting a system to analyze each song's spectrum into short samples and vote for the results. Among Deep Learning algorithms, CNN showed superior performance in the category of music genre compared to RNN, and improved performance when CNN and RNN were applied together. The system of voting for each CNN result by Deep Learning a short sample of music showed better results than the previous model and the model with Softmax layer added to the model performed best. The need for the explosive growth of digital media and the automatic classification of music genres in numerous streaming services is increasing. Future research will need to reduce the proportion of undifferentiated songs and develop algorithms for the last category classification of undivided songs. [ABSTRACT FROM AUTHOR]