• 한국시뮬레이션학회논문지
• /
• 제31권1호
• /
• pp.11-18
• /
• 2022

최근 공공장소 및 시설에서 범죄예방 및 시설 안전을 목적으로 영상정보 기반의 인체의 행위를 분류하는 연구가 활발히 진행되고 있다. 이러한 인체 행위분류의 성능을 향상하기 위해서 대부분의 연구는 전이학습 기반의 딥러닝을 적용하고 있다. 그러나 딥러닝의 기반이 되는 중추 네트워크 모델(Backbone Network Model)의 수가 증가하고 아키텍처가 다양해짐에도 불구하고, 소수의 모델만 사용하는 분위기 때문에 운용목적에 적합한 중추 네트워크 모델을 찾는 연구는 미흡한 실정이다. 본 연구는 영상정보를 기초로 인체 행위를 분류하는 인공지능 모델을 개발하기 위해 최근에 개발된 5가지의 딥러닝 중추 네트워크 모델을 대상으로 전이학습을 적용하고 각 모델의 정확도 및 학습효율 측면에서 비교 및 분석하여 가장 효율이 높은 모델을 제안하였다. 이를 위해, 기본적인 인체 행위가 아닌 운동 종목 기반의 활동적이고 신체접촉이 높은 12가지의 인체 활동을 선정하고 관련된 7,200개의 이미지를 수집하였으며, 5가지의 중추 네트워크 모델에 총 20회의 전이학습을 균등하게 적용하고 학습과정과 결과성능을 통해 인체 행위를 분류하는데 적합한 중추 네트워크 모델을 정량적으로 비교 및 분석하였다. 그 결과 XceptionNet 모델이 학습 및 검증 정확도에서 0.99 및 0.91로, Top 2 및 평균 정밀도에서 0.96 및 0.91로 나타났으며 학습 소요시간은 1,566초, 모델용량의 크기는 260.4MB로 정확도와 학습효율 측면에서 다른 모델보다 높은 성능이 나타남을 확인할 수 있었다. 이러한 결과는 전이학습을 적용하여 인체 행위분류를 진행하는 다양한 연구 분야에 활용되기를 기대한다.

• 인터넷정보학회논문지
• /
• 제22권3호
• /
• pp.9-16
• /
• 2021

This study aims to develop a human activity recognition (HAR) system as a Deep-Learning (DL) classification model, distinguishing various human activities. We solely rely on the signals from a wristband accelerometer worn by a person for the user's convenience. 3-axis sequential acceleration signal data are gathered within a predefined time-window-slice, and they are used as input to the classification system. We are particularly interested in developing a Deep-Learning model that can outperform conventional machine learning classification performance. A total of 13 activities based on the laboratory experiments' data are used for the initial performance comparison. We have improved classification performance using the Convolutional Neural Network (CNN) combined with an auto-encoder feature reduction and parameter tuning. With various publically available HAR datasets, we could also achieve significant improvement in HAR classification. Our CNN model is also compared against Recurrent-Neural-Network(RNN) with Long Short-Term Memory(LSTM) to demonstrate its superiority. Noticeably, our model could distinguish both general activities and near-identical activities such as sitting down on the chair and floor, with almost perfect classification accuracy.

• 한국멀티미디어학회논문지
• /
• 제21권12호
• /
• pp.1459-1466
• /
• 2018

Robots which detect human and recognize action are important factors for human interaction, and many researches have been conducted. Recently, deep learning technology has developed and learning based robot's technology is a major research area. These studies require a database to learn and evaluate for intelligent human perception. In this paper, we propose a multi-modal sensor-based image database condition considering the security task by analyzing the image database to detect the person in the outdoor environment and to recognize the behavior during the running of the robot.

• East Asian mathematical journal
• /
• 제29권2호
• /
• pp.137-168
• /
• 2013

In this paper we study activity theory and Davydov's learning activity theory. We analyze brief history of activity theory in Russia, structure of human activity, and Davydov's studies in activity theory. Especially we analyze Davydov's 1st grade mathematics textbook, and try to investigate embodiment of Davydov's learning activity theory in his mathematics textbook.

• 인터넷정보학회논문지
• /
• 제21권4호
• /
• pp.127-135
• /
• 2020

The purpose of this study is to develop a human activity recognition (HAR) system, which distinguishes 13 activities, including five activities commonly dealt with in conventional HAR researches and eight activities from the Eastern-Asian culture. The eight special activities include floor-sitting/standing, chair-sitting/standing, floor-lying/up, and bed-lying/up. We used a 3-axis accelerometer sensor on the wrist for data collection and designed a machine learning model for the activity classification. Data clustering through preprocessing and feature extraction/reduction is performed. We then tested six machine learning algorithms for recognition accuracy comparison. As a result, we have achieved an average accuracy of 99.7% for the 13 activities. This result is far better than the average accuracy of current HAR researches based on a smartwatch (89.4%). The superiority of the HAR system developed in this study is proven because we have achieved 98.7% accuracy with publically available 'pamap2' dataset of 12 activities, whose conventionally met the best accuracy is 96.6%.

• 한국인터넷방송통신학회논문지
• /
• 제22권5호
• /
• pp.177-183
• /
• 2022

최근 COVID-19로 인해 홈 트레이닝의 관심도가 증가하고 있다. 이에 따라 HAR(human activity recognition) 기술을 홈 트레이닝에 적용한 연구가 진행되고 있다. 기존 HAR 분야의 논문에서는 동적인 자세보다는 앉기, 일어서기와 같은 정적인 자세들을 분석한다. 본 논문은 동적인 운동 자세를 분석하여 사용자의 운동 자세 정확도를 보여주는 딥러닝 모델을 제안한다. AI hub의 피트니스 이미지를 blaze pose를 사용하여 사람의 자세 데이터를 분석한다. 3개의 딥러닝 모델: RNN(recurrnet neural networks), LSTM(long short-term memory networks), CNN(convolution neural networks)에 대하여 실험을 진행한다. RNN, LSTM, CNN 모델의 f1-score는 각각 0.49, 0.87, 0.98로 CNN 모델이 가장 적합하다는 것을 확인하였다. 이후 연구로는, 다양한 학습 데이터를 사용하여 더 많은 운동 자세를 분석할 예정이다.

• Journal of Electrical Engineering and Technology
• /
• 제13권2호
• /
• pp.961-968
• /
• 2018

Recently, researches on automatic recognition of human activities have been actively carried out with the emergence of various intelligent systems. Since a large amount of visual data can be secured through Closed Circuit Television, it is required to recognize human behavior in a dynamic situation rather than a static situation. In this paper, we propose new intelligent human activity recognition model using the trajectory information extracted from the video sequence. The proposed model consists of three steps: segmentation and partitioning of trajectory step, feature extraction step, and behavioral learning step. First, the entire trajectory is fuzzy partitioned according to the motion characteristics, and then temporal features and spatial features are extracted. Using the extracted features, four pedestrian behaviors were modeled by decision tree learning algorithm and performance evaluation was performed. The experiments in this paper were conducted using Caviar data sets. Experimental results show that trajectory provides good activity recognition accuracy by extracting instantaneous property and distinctive regional property.

• 한국멀티미디어학회논문지
• /
• 제20권8호
• /
• pp.1421-1429
• /
• 2017

Human activity recognition has been studied using various sensors and algorithms. Human activity recognition can be divided into sensor based and vision based on the method. In this paper, we proposed an activity recognition system using acceleration sensor and gyroscope sensor in smartphone among sensor based methods. We used Deep Belief Network (DBN), which is one of the most popular deep learning methods, to improve an accuracy of human activity recognition. DBN uses the entire input set as a common input. However, because of the characteristics of different time window depending on the type of human activity, the RBMs, which is a component of DBN, are configured hierarchically by combining them from different time windows. As a result of applying to real data, The proposed human activity recognition system showed stable precision.

• 대한임베디드공학회논문지
• /
• 제18권6호
• /
• pp.319-325
• /
• 2023

In this study, we proposed a method for hyperparameter optimization in the building and training of a deep learning model designed to process point cloud data collected by a millimeter-wave radar system. The primary aim of this study is to facilitate the deployment of a baseline model in resource-constrained IoT devices. We evaluated a RadHAR baseline deep learning model trained on a public dataset composed of point clouds representing five distinct human activities. Additionally, we introduced a coarse-to-fine hyperparameter optimization procedure, showing substantial potential to enhance model efficiency without compromising predictive performance. Experimental results show the feasibility of significantly reducing model size without adversely impacting performance. Specifically, the optimized model demonstrated a 3.3% improvement in classification accuracy despite a 16.8% reduction in number of parameters compared th the baseline model. In conclusion, this research offers valuable insights for the development of deep learning models for resource-constrained IoT devices, underscoring the potential of hyperparameter optimization and model size reduction strategies. This work contributes to enhancing the practicality and usability of deep learning models in real-world environments, where high levels of accuracy and efficiency in data processing and classification tasks are required.

• 전자통신동향분석
• /
• 제33권3호
• /
• pp.89-99
• /
• 2018

Human activity recognition (HAR) is a technology that aims to offer an automatic recognition of what a person is doing with respect to their body motion and gestures. HAR is essential in many applications such as human-computer interaction, health care, rehabilitation engineering, video surveillance, and artificial intelligence. Smartphones are becoming the most popular platform for activity recognition owing to their convenience, portability, and ease of use. The noticeable change in smartphone-based activity recognition is the adoption of a deep learning algorithm leading to successful learning outcomes. In this article, we analyze the technology trend of activity recognition using smartphone sensors, challenging issues for future development, and a strategy change in terms of the generation of a activity recognition dataset.