• The Journal of the Convergence on Culture Technology
• /
• v.7 no.2
• /
• pp.411-417
• /
• 2021

The field that studies human-computer interaction is called HCI (Human-computer interaction). This field is an academic field that studies how humans and computers communicate with each other and recognize information. This study is a study on hand gesture recognition for human interaction. This study examines the problems of existing recognition methods and proposes an algorithm to improve the recognition rate. The hand region is extracted based on skin color information for the image containing the shape of the human hand, and the center of gravity profile is calculated using principal component analysis. I proposed a method to increase the recognition rate of hand gestures by comparing the obtained information with predefined shapes. We proposed a method to increase the recognition rate of hand gestures by comparing the obtained information with predefined shapes. The existing center of gravity profile has shown the result of incorrect hand gesture recognition for the deformation of the hand due to rotation, but in this study, the center of gravity profile is used and the point where the distance between the points of all contours and the center of gravity is the longest is the starting point. Thus, a robust algorithm was proposed by re-improving the center of gravity profile. No gloves or special markers attached to the sensor are used for hand gesture recognition, and a separate blue screen is not installed. For this result, find the feature vector at the nearest distance to solve the misrecognition, and obtain an appropriate threshold to distinguish between success and failure.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.18 no.6
• /
• pp.1009-1014
• /
• 2023

Comprehensive observations of the Euler method and the Lagrangian method were performed in order to obtain high-resolution observation data in space and time for the complex environment of new city. The two radiosondes, which measure meteorological parameters using Lagrangian methods, produced air pressure, wind speed and wind direction. They were generally consistent with each other even if the observation points or times were different. The temperature measured by the sensor exposed to the air during the day was relatively high as the altitude increased due to the influence of solar radiation. The temporal difference in wind direction and speed was found in the comparison of Euler's wind profiler data with radiosonde data. When the wind field is horizontally in homogeneous, this result implies the need to consider the advection component to compare the data of the two observation methods. In this study, a method of using observation data at different times for each altitude section depending on the observation period of the Euler method is proposed to effectively compare the data of the two observation methods.

• Korean Journal of Remote Sensing
• /
• v.37 no.6_1
• /
• pp.1631-1645
• /
• 2021

This research aims to provide the characteristics of the world's first active lidar sensor Atmospheric Laser Doppler Instrument (ALADIN) wind data and Geostationary Korea Multi Purpose Satellite 2A (GK2A) Atmospheric Motion Vector (AMV) data by comparing two wind data. As a result of comparing the data from September 2019 to August 1, 2020, The total number of collocated data for the AMV (using IR channel) and Mie channel ALADIN data is 177,681 which gives the Root Mean Square Error (RMSE) of 3.73 m/s and the correlation coefficient is 0.98. For a more detailed analysis, Comparison result considering altitude and latitude, the Normalized Root Mean Squared Error (NRMSE) is 0.2-0.3 at most latitude bands. However, the upper and middle layers in the lower latitudes and the lower layer in the southern hemispheric are larger than 0.4 at specific latitudes. These results are the same for the water vapor channel and the visible channel regardless of the season, and the channel-specific and seasonal characteristics do not appear prominently. Furthermore, as a result of analyzing the distribution of clouds in the latitude band with a large difference between the two wind data, Cirrus or cumulus clouds, which can lower the accuracy of height assignment of AMV, are distributed more than at other latitude bands. Accordingly, it is suggested that ALADIN wind data in the southern hemisphere and low latitude band, where the error of the AMV is large, can have a positive effect on the numerical forecast model.

• Journal of Intelligence and Information Systems
• /
• v.25 no.1
• /
• pp.163-177
• /
• 2019

As smartphones are getting widely used, human activity recognition (HAR) tasks for recognizing personal activities of smartphone users with multimodal data have been actively studied recently. The research area is expanding from the recognition of the simple body movement of an individual user to the recognition of low-level behavior and high-level behavior. However, HAR tasks for recognizing interaction behavior with other people, such as whether the user is accompanying or communicating with someone else, have gotten less attention so far. And previous research for recognizing interaction behavior has usually depended on audio, Bluetooth, and Wi-Fi sensors, which are vulnerable to privacy issues and require much time to collect enough data. Whereas physical sensors including accelerometer, magnetic field and gyroscope sensors are less vulnerable to privacy issues and can collect a large amount of data within a short time. In this paper, a method for detecting accompanying status based on deep learning model by only using multimodal physical sensor data, such as an accelerometer, magnetic field and gyroscope, was proposed. The accompanying status was defined as a redefinition of a part of the user interaction behavior, including whether the user is accompanying with an acquaintance at a close distance and the user is actively communicating with the acquaintance. A framework based on convolutional neural networks (CNN) and long short-term memory (LSTM) recurrent networks for classifying accompanying and conversation was proposed. First, a data preprocessing method which consists of time synchronization of multimodal data from different physical sensors, data normalization and sequence data generation was introduced. We applied the nearest interpolation to synchronize the time of collected data from different sensors. Normalization was performed for each x, y, z axis value of the sensor data, and the sequence data was generated according to the sliding window method. Then, the sequence data became the input for CNN, where feature maps representing local dependencies of the original sequence are extracted. The CNN consisted of 3 convolutional layers and did not have a pooling layer to maintain the temporal information of the sequence data. Next, LSTM recurrent networks received the feature maps, learned long-term dependencies from them and extracted features. The LSTM recurrent networks consisted of two layers, each with 128 cells. Finally, the extracted features were used for classification by softmax classifier. The loss function of the model was cross entropy function and the weights of the model were randomly initialized on a normal distribution with an average of 0 and a standard deviation of 0.1. The model was trained using adaptive moment estimation (ADAM) optimization algorithm and the mini batch size was set to 128. We applied dropout to input values of the LSTM recurrent networks to prevent overfitting. The initial learning rate was set to 0.001, and it decreased exponentially by 0.99 at the end of each epoch training. An Android smartphone application was developed and released to collect data. We collected smartphone data for a total of 18 subjects. Using the data, the model classified accompanying and conversation by 98.74% and 98.83% accuracy each. Both the F1 score and accuracy of the model were higher than the F1 score and accuracy of the majority vote classifier, support vector machine, and deep recurrent neural network. In the future research, we will focus on more rigorous multimodal sensor data synchronization methods that minimize the time stamp differences. In addition, we will further study transfer learning method that enables transfer of trained models tailored to the training data to the evaluation data that follows a different distribution. It is expected that a model capable of exhibiting robust recognition performance against changes in data that is not considered in the model learning stage will be obtained.