Mobile QR Code QR CODE

Sentiment Analysis using Various Machine Learning Techniques: A Review

(Pinky Yadav) ; (Madhumita Kathuria)

Sentiment analysis is the study of deriving opinions and insights from a tremendous amount of text data present on various social media platforms like Facebook, Twitter, and many others. Nowadays, it is vital for companies and organizations to behold the reviews or public opinion about their product to survive in a competitive market. In sentiment analysis, we categorize this data into nominal labels or polarity, which maps the review to an overall sentiment. This research paper provides a review of various machine learning techniques for sentiment analysis in detail on social media data. Along with the text reviews, the emoticons between the text will also be analyzed to achieve the exact sentiment of the reviewer.

No-reference Image Contrast Quality Assessment based on the Degree of Uniformity in Probability Distribution

(Wonvin Kim) ; (Changhoon Yim)

We propose a novel contrast-based no-reference image quality assessment (NR-IQA) method. The proposed method quantifies the degree of uniformity in the probability distribution of the pixel values of an image to measure contrast quality without the need for a reference image. To quantify the degree of symmetric uniformity in the distribution of pixel values for image contrast assessment, we divide the intensity levels into left and right ranges recursively at each depth as a binary tree. This symmetric hierarchical recursive calculation of the degree of uniformity for contrast measurement provides a good metric for NR-IQA. Experimental results show that the proposed method is more consistent with the levels of contrast decrement and is superior in assessing contrast-enhanced and dehazed images compared with state-of-the-art and classic NR-IQA methods. The proposed method is more suited to real-time applications due to its lower computational complexity compared with previous methods.

A 1D CNN-LSTM using Wav2vec 2.0 for Violent Scene Discrimination

(Huiyong Bak) ; (Sangmin Lee)

In this paper, an effective system for discriminating violent scenes in movies from audio signals alone is proposed. The technology for automatic discrimination of violent scenes is one of the most crucial aspects of media filtering, protecting users from undesired media. Previous studies have conducted violent scene discrimination using a mel spectrogram and 2D convolutional neural networks (CNNs); however, the mel spectrogram cannot extract mutual information from audio, and 2D CNNs are unsuitable for audio. Therefore, these models do not yield good performance. The system proposed in this paper extracts audio features by using Wav2vec 2.0, which can extract mutual information from audio. The features of the extracted audio are inputted to a 1D CNN and long short-term memory (LSTM), which are algorithms suitable for audio, and violent scenes are discriminated through fully connected and softmax layers. To evaluate the proposed system, violent scenes are discriminated using the Violent Movie Scenes Dataset (VMD). As a result, the accuracy of the proposed system when discriminating violent scenes is 96.25%, providing better performance than in previous studies.

Recognition of Respiratory Instability using a Photoplethysmography of Wrist-watch type Wearable Device

(Seungyoon Nam) ; (John Lorenzo Bautista) ; (Chanyoung Hahm) ; (Hyunsoon Shin)

The photoplethysmography signal is composed of a cardiac-synchronous pulsatile waveform and different parts, which is modulated in amplitude by respiration. This paper presents a new indexing method similar to the apnea-hypopnea index and respiratory disturbance index for the self-diagnosis of sleep apnea symptoms (central and obstructed apnea) by using only a photoplethysmogram (PPG) signal. Sleep apnea is a sleeping disorder from several chronic conditions in which partial or complete cessation of breathing occurs many times throughout sleep at night. A respiratory rate signal (respiration-induced intensity variation) is modulated by synchronizing with the breathing rhythm extracted from PPG using a reflected light on the top of the wrist. This paper presents a new automated recognition and estimation method for daytime apnea and sleep-induced apnea using a wristwatch-type wearable device that can recognize irregular breathing using respiratory rate frequency-based features. The new respiratory effort strength index is proposed to quantify sleep apnea by determining how much a patient is suffering.

Where to Look: Visual Attention Estimation in Road Scene Video for Safe Driving

(Yeejin Lee) ; (Byeongkeun Kang)

This work addresses the task of locating regions that are more crucial for safe driving than other areas on roads. It could be utilized to improve the efficiency and safety of autonomous driving vehicles or robots and could also be useful for human drivers when employed in driver-assistance systems. To achieve robust and accurate attention prediction, we propose a multiscale color and motion-based attention prediction network. The network consists of three components where each processes multi-scaled color images, uses multi-scaled motion information, and merges the outputs of the two streams, respectively. The proposed network is guided to utilize the movement of objects/people as well as the type/location of things/stuff. We demonstrate the effectiveness of the proposed system by experimenting with an actual driving dataset. The experimental results show that the proposed framework outperforms previous works.

Local Non-linear Quantization for Neural Network Compression in MPEG-NNR

(Hyeon-Cheol Moon) ; (Jae-Gon Kim)

Deep Convolutional Neural Networks (CNNs) have demonstrated excellent performance in various visual applications, but their deployment, especially in resource-constrained environments, is limited due to their enormous computational complexity and memory requirements. Therefore, compression of network models while still maintaining the task performance of the trained model is being studied. Recently, the Moving Picture Experts Group (MPEG) developed a standard called Neural Network compression and Representation (NNR) that provides a compressed representation of trained neural networks in an interoperable form. In this paper, we propose a local non-linear quantization (LNQ) method for compressing weight parameters of neural network models. The experimental results show that the proposed LNQ achieves about a 29% gain in compression efficiency with virtually no loss of performance in the tasks, compared to the NNR test model, called the Neural network Compression Test Model (NCTM) version 3.0.

Visual SLAM-based Vehicle Control for Autonomous Valet Parking

(Younggon Jo) ; (Seokhyeon Hong) ; (Jeongmok Ha) ; (Sungsoo Hwang)

This research proposes an efficient vehicle control method using visual SLAM (Simultaneous Localization And Mapping) for AVP (Autonomous Valet Parking). SLAM technology generates a map of the surrounding environment and localizes the vehicle within the map. It is used to identify the layout of the parking lot and track the vehicle by using camera sensors only. In the proposed system, an autonomous driving vehicle is controlled using the coordinates of the keyframe on the visual SLAM map. The vehicle is driven by determining the keyframe in the movable position during the autonomous driving process. This driving procedure is possible because the coordinates of the vehicle and the keyframe can be estimated through the SLAM map. However, the SLAM map, generated using features of the surrounding environment, is likely to change scale while driving due to feature matching errors. Therefore, the system proposes to update the initial scale using the time the vehicle has moved and the changes in vehicle coordinates on the SLAM map. The tracking success rate of autonomous driving and the success rate of autonomous parking were measured to evaluate the performance of the proposed system. The experimental results indicate that autonomous valet parking can be achieved using visual SLAM.

Blind Image Deblurring based on Deep Image Prior

(Changwoo Lee) ; (Jinwon Choi)

Many studies on image deblurring have been conducted, and deep learning methods for blind image deblurring have received considerable attention due to their good performance. Recently, the SelfDeblur method was proposed for blind image deblurring based on deep image prior (DIP). In the SelfDeblur method, two neural networks for an image generator and a blur kernel generator are learned simultaneously with only one blurry image. This shows the feasibility of blind image deblurring using unsupervised learning, since it requires no training process. In this paper, we propose a method to maximize the performance of blind image deblurring based on DIP. The optimal loss function for deep learning is studied for the SelfDeblur method, and the deblurring performance of the proposed method is stabilized and maximized using the image prior and the kernel prior for the total loss function. Extensive computer simulations show that the proposed method yields superior performance compared to conventional methods.

Space Division Multiple Access Base Station (SDMA) Based on Block Adaptive Euclidean Direction Search Algorithm

(Vian S. AL-Doori) ; (Thamer M. Jamel) ; (Bashar M. Mansoor)

An impulse response is present in long-length wireless communication channels in communication systems such as massive MIMO or 5G systems that use block coding, such as Low-Density Parity-Check (LDPC) code. To deal with it, it is favorable to use block adaptive filtering instead of sequential or traditional algorithms. In this study, a block adaptive algorithm called the Speedy Euclidean Direction Search (SEDS) algorithm is proposed for a Space Division Multiple Access Base Station (SDMA) to perform adaptive beamforming operation of the system. An investigation and analysis of the performance of the SEDS algorithm were conducted using SDMA with a Multiple Input Single Output (MISO) downlink scheme that uses block processing of the data samples for a single-cell downlink. Moreover, a novel and fair comparative study was carried out between SEDS and the Block adaptive Least Mean Square (BLMS) algorithm. As far as we know, no research has used the BLMS algorithm for the SDMA system, which is another contribution of this study. The simulation results demonstrate that the SEDS algorithm has fast convergence and a very accurate estimation compared with BLMS. Moreover, the SEDS algorithm has better performance in indoor environments compared with BLMS. On the other hand, SEDS starts to be unstable for a large value of block length (L), whereas BLMS stays stable especially for outdoor environments.

A Light-Weight AES Design using LFSR-based S-Box for IoT Applications

(Donghui Lee) ; (Myeongjin Kwak) ; (Jungwon Lee) ; (Beomjun Kim) ; (Yongtae Kim)

This paper presents a novel light-weight AES architecture based on linear feedback shift register (LFSR). Since the traditional substitution-box (S-box) of the AES architecture consumes a large amount of hardware resources, we propose LFSR-based S-box and inverse S-box designs to reduce the hardware overheads significantly. In addition, we replace the conventional AES key scheduler with the LFSR-based round key generator to improve the hardware efficiency further. When implemented in a 32-nm CMOS technology, the proposed designs improve the area, delay, power, and energy by 57.4%, 64.3%, 30.1%, and 75% in the AES-128, 52.2%, 12.5%, 21.9%, and 31.7% in the AES-192, and 54.8%, 5.2%, 23.4%, and 27.4% in the AES-256, respectively, when compared to the traditional S-box based AES architecture. In addition, a joint analysis of our designs in terms of area, delay, and energy shows that the proposed AES designs enhance the area-delay product (ADP) and energy-delay product (EDP) by up to 84.7% and 82.5%, respectively, compared to the traditional design.