• Radiation Oncology Journal
• /
• v.15 no.4
• /
• pp.393-402
• /
• 1997

Purpose : When an x-ray beam of small field size is irradiated to target area containing an air cavity, such as larynx, the underdosing effect is observed in the region near the interfaces of air and soft tissue. With a larynx model, air cavity embedded in tissue-equivalent material, this study is intonded for examining Parameters, such as beam quality, field size, and cavity size, to affect the dose distribution near the air cavity. Materials and Methods : Three x-rar beams, 4-, 6- and 10-MV, were employed to Perform a measurement using a 2cm $(width){\times}L$ (length in cm, one side of x-ray field used 2cm (height) air cavity in the simulated larynx. A thin window parallel-plate chamber connected to an electrometer was used for a dosimetry system. A ratio of the dose at various distances from the cavity-tissue interface to the dose at the same points in a homogeneous Phantom (ebservedlexpected ratio, O/E) normalized buildup curves, and ratio of distal surface dose to dose at the maximum buildup depth were examined for various field sizes. Measurement for cavity size effect was performed by varying the height (Z) of the air cavity with the width kept constant for several field sizes. Results : No underdosing effect for 4-MV beam for fields larger than $5cm\times5cm$ was found For both 6- and 10-MV beams, the underdosing portion of the larynx at the distal surface was seen to occur for small fields, $4cm\times4cm\;and\;5cm\times5cm$. The underdosed tissue was increased in its volume with beam energy even for similar surface doses. The relative distal surface dose to maximum dose was changed to 0.99 from 0.95, 0.92, and 0.91 for 4-, 6-, and 10-MV, respectively, with increasing field size, $4cm\times4cm\;to\;8cm\times8cm$, For 6- and 10-MV beams, the dose at the surface of the cavity is measured less than the predicted by about two and three percent. respectively. but decrease was found for 4-MV beam for $5cm\times5cm$ field. For the $4cm\timesL\timesZ$ (height in cm). varying depth from 0.0 to 4.8cm, cavity, O/E> 1.0 was observed regardless of the cavity size for any field larger than about $8cm\times8cm$. Conclusion : The magnitude of underdosing depends on beam energy, field size. and cavity size for the larynx model. Based on the result of the study. caution must be used when a small field of a high quality x-ray beam is irradiated to regions including air cavities. and especially the region where the tumor extends to the surface. Low quality beam. such as. 4-MV x-ray, and larger fields can be used preferably to reduce the risk of underdosing, local failure. In the case of high quality beams such as 6- and 10-MV x-rays, however. an additional boost field is recommended to add for the compensation of the underdosing region when a typically used treatment field. $8cm\times8cm$, is employed.

• Radiation Oncology Journal
• /
• v.26 no.4
• /
• pp.204-212
• /
• 2008

Purpose: This study was performed to evaluate the disease-free survival and risk factors of recurrence in early breast cancer patients who have undergone breast conserving surgery and radiation therapy. Materials and Methods: From March 1997 to December 2002, 77 breast cancer patients who underwent breast conserving surgery and radiation therapy were reviewed retrospectively. The median follow-up time was 58.4 months (range $43.8{\sim}129.4$ months) and the mean subject age was 41 years. The frequency distribution of the different T stages, based on the tumor characteristics was 38 (49.3%) for T1, 28 (36.3%) for T2, 3 for T3, 7 for T is and 1 for an unidentified sized tumor. In addition, 52 patients (67.5%) did not have axillary lymph metastasis, whereas 14 patients (18.1%) had $1{\sim}3$ lymph node metastases and 3 (0.03%) had more than 4 lymph node metastases. The resection margin was negative in 59 patients, close (${\leq}2\;mm$) in 15, and positive in 4. All patients received radiation therapy at the intact breast using tangential fields with a subsequent electron beam boost to the tumor bed at a total dose ranging from 59.4 Gy to 66.4 Gy. Patients with more than four positive axillary lymph nodes received radiation therapy ($41.4{\sim}60.4\;Gy$) at the axillary and supraclavicular area. Chemotherapy was administered in 59 patients and tamoxifen or fareston was administered in 29 patients. Results: The 5 year overall survival and disease-free survival rates were 98.08% and 93.49%, respectively. Of the 77 patients, a total of 4 relapses (5.2%), including 1 isolated supraclavicular relapse, 1 supraclavicular relapse with synchronous multiple distant relapses, and 2 distant relapses were observed. No cases of local breast relapses were observed. Lymph node metastasis or number of metastatic lymph nodes was not found to be statistically related with a relapse (p=0.3289) nor disease-free survival (p=0.1430). Patients with positive margins had a significantly shorter disease-free survival period (p<0.0001) and higher relapse rates (p=0.0507). However, patients with close margins were at equal risk of relapse and disease-free survival as with negative margins (p=1.000). Patients younger than 40 years of age had higher relapse rates (9.3% vs. 0%) and lower disease-free survival periods, but the difference was not statistically significant (p=0.1255). The relapse rates for patients with tumors was 14% for tumor stage T2, compared to 0% for tumor stage T1 tumors (p=0.0284). A univariate analysis found that disease-free survival and relapse rates, T stage, positive resection margin and mutation of p53 were significant factors for clinical outcome. Conclusion: The results of this study have shown that breast conservation surgery and radiation therapy in early breast cancer patients has proven to be a safe treatment modality with a low relapse rate and high disease-free survival rate. The patients with a positive margin, T2 stage, and mutation of p53 are associated with statistically higher relapse rates and lower disease-free survival.

• Journal of the Korean Institute of Traditional Landscape Architecture
• /
• v.35 no.2
• /
• pp.26-44
• /
• 2017

Regarded as Korea's traditional pond, Gungnamj Pond was surmised to be "Gungnamji" due to its geological positioning in the south of Hwajisan (花枝山) and relics of the Gwanbuk-ri (官北里) suspected of being components to the historical records of Muwang (武王)'s pond of The Chronicles of the Three States [三國史記] and Sabi Palace, respectively, yet was subjected to a restoration following a designation to national historic site. This study is focused on the distortion of authenticity identified in the course of the "Gungnamji Pond" restoration and the invention of tradition, whose summarized conclusions are as follows. 1. Once called Maraebangjuk (마래방죽), or Macheonji (馬川池) Pond, Gungnamji Pond was existent in the form of a low-level swamp of vast area encompassing 30,000 pyeong during the Japanese colonial period. Hong, Sa-jun, who played a leading role in the restoration of "Gungnamji Pond," said that even during the 1940s, the remains of the island and stone facilities suspected of being the relics of Gungnamji Pond of the Baekje period were found, and that the traces of forming a royal palace and garden were discovered on top of them. Hong, Sa-jun also expressed an opinion of establishing a parallel between "Gungnamji Pond" and "Maraebangjuk" in connection with a 'tale of Seodong [薯童說話]' in the aftermath of the detached palace of Hwajisan, which ultimately operated as a theoretical ground for the restoration of Gungnamj Pond. Assessing through Hong, Sa-jun's sketch, the form and scale of Maraebangjuk were visible, of which the form was in close proximity to that photographed during the Japanese colonial period. 2. The minimized restoration of Gungnamji Pond faced deterrence for the land redevelopment project implemented in the 1960s, and the remainder of the land size is an attestment. The fundamental problem manifest in the restoration of Gungnamji Pond numerously attempted from 1964 through 1967 was the failure of basing the restorative work in the archaeological facts yet in the perspective of the latest generations, ultimately yielding a replication of Hyangwonji Pond of Gyeongbok Palace. More specifically, the methodologies employed in setting an island and a pavilion within a pond, or bridging an island with a land evidenced as to how Gungnamji Pond was modeled after Hyangwonji Pond of Gyeongbok Palace. Furthermore, Chihyanggyo (醉香橋) Bridge referenced in the designing of the bridge was hardly conceived as a form indigenous to the Joseon Dynasty, whose motivation and idea of the misguided restoration design at the time all the more devaluated Gungnamji Pond. Such an utterly pure replication of the design widely known as an ingredient for the traditional landscape was purposive towards the aesthetic symbolism and preference retained by Gyeongbok Palace, which was intended to entitle Gungnamji Pond to a physical status of the value in par with that of Gyeongbok Palace. 3. For its detachment to the authenticity as a historical site since its origin, Gungnamji Pond represented distortions of the landscape beauty and tradition even through the restorative process. The restorative process for such a historical monument, devoid of constructive use and certain of distortion, maintains extreme intimacy with the nationalistic cultural policy promoted by the Park, Jeong-hee regime through the 1960s and 1970s. In the context of the "manipulated discussions of tradition," the Park's cultural policy transformed the citizens' recollection into an idealized form of the past, further magnifying it at best. Consequently, many of the historical sites emerged as fancy and grand as they possibly could beyond their status quo across the nation, and "Gungnamji Pond" was a victim to this monopolistic government-led cultural policy incrementally sweeping away with new buildings and structures instituted regardless of their original space, and hence, their value.

• Journal of agricultural medicine and community health
• /
• v.26 no.1
• /
• pp.81-103
• /
• 2001

This study was conducted to identify the health behavior of middle-aged rural women and the factors that have an effect on them. For the purpose of the study, examinations were made from March 01, 2000 to March 31, 2000 with 468 women aged 40 to 64 out of 2,263 people whom four Primary Health Posts located in Yechon County, Kyongsangbuk-do Province, are in charge of. The results are summarized as follows. 17.5% of the subjects responded that the extent of their own interest in health were high. For the subjects having a chronic disease, a nuclear family, or an open family atmosphere, the extent appeared to be relatively higher, 15.4% responded that the extent of family's interest in their health was high. It was significantly high if the extent of education was high or if the family atmosphere was open. The subjects' average score of self-efficacy was 49.9 out of 68. The score significantly varied depending on religion, education, living together with a spouse or not, and the extent of the subjects' interest in health. The family pattern, family atmosphere, family's interest in the subjects' health were the variables that significantly influenced the self-efficacy. The average score of family function was 5.51 out of 10. The score significantly varies depending on age, education, occupation, financial status, the extent of the subjects' own interest in health, family atmosphere and family's interest in the subjects' health. In the practice of health behavior, the nonsmoking rate was 89.5%, the nondrinking rate 63.0%, the rate of exercising practice 6.6%, the rate of normal sleeping 75.6%, the rate of eating breakfast 91.7%, the rate of not eating between meals 18.2%, and the standard BMI 69.2%. In the frequency of health behavior, the subjects with the Breslow Index of 0-3, 4-5 and 6-7 accounted for 4.5%, 53.2%, and 42.3%, respectively. The average score of health behavior was 5.20 out of 7, in which significant variables were living together with a spouse or not, financial status, absence or presence of a chronic disease, and family atmosphere. In the multiple regression analysis with health behavior as a dependent variable, it was shown that living together with a spouse or not, financial status, and family atmosphere were the significantly substantial variables. The subjects were found to do health behavior well if they had not a spouse, a good financial status, or an open family atmosphere. They were also found to do health behavior well if the extent of self-efficacy was high or if the extent of family function was low, but these were not the significant variables. It is needed to develop a standard measuring tool fit for our environment and perform more studies in the future because the measuring tool used in this study was a tool developed in a foreign county. In promoting community health projects, it is required not to provide all community people with a uniform health program but to identify the health behavior of individuals and other variables such as living together with a spouse or not, financial status and family atmosphere before arranging for a proper health program.

• Korea journal of population studies
• /
• v.19 no.2
• /
• pp.77-105
• /
• 1996

The rural problems which we are facing start from the extremely small sized population and the skewed population structure by age and sex. Thus we analyzed the change of the rural population. And we analyzed the recent return migration to the rural areas by comparing the recent in-migrants with out-migrants to rural areas. And by analyzing the rural village survey data which was to show the current characteristics of rural population, we found out the effects of the in-migrants to the rural areas and predicted the futures of rural villages by characteristics. The changes of rural population composition by age was very clear. As the out-migrants towards cities carried on, the population composition of young children aged 0~4 years was low and the aged became thick. The proportion of the population aged 0~4 years was 45.1% of the total population in 1970 and dropped down to 20.4% in 1995, which is predicted to become under 20% from now on. In the same period(1970~1995), the population aged 65 years and over rose from 4.2% to 11.9%. In 1960, before industrialization, the proportion of the population aged 0~4 years in rural areas was higher than that of cities. As the rural young population continuously moves to cities it became lower than that in urban areas from 1975 and the gap grew till 1990. But the proportion of rural population aged 0~4 years in 1995 became 6.2% and the gap reduced. We can say this is the change of the characteristics of in-migrants and out-migrants in the rural areas. Also considering the composition of the population by age group moving from urban to rural area in the late 1980s, 51.8% of the total migrants concentrates upon age group of 20~34 years and these people's educational level was higher than that of out-migrants to urban areas. This fact predicted the changes of the rural population, and the results will turn out as a change in the rural society. However, after comparing the population structure between the pure rural village of Boeun-gun and suburban village of Paju-gun which was agriculture centered village but recently changed rapidly, the recent change of the rural population structure which the in-migrants to rural areas becomes younger is just a phenomenon in the suburban rural areas, not the change of the total rural areas in general. From the characteristics of the population structure of rural village from the field survey on these villages, we can see that in the pure rural villages without any effects from cities the regidents are highly aged, while industrialization and urbanization are making a progress in suburban villages. Therefore, the recent partial change of the rural population structure and the change of characteristics of the in-migrants toward rural areas is effecting and being effected by the population change of areas like suburban rural villages. Although there are return migrants to rural areas to change their jobs into agriculture, this is too minor to appear as a statistic effect.

• Journal of Intelligence and Information Systems
• /
• v.20 no.1
• /
• pp.35-48
• /
• 2014

According to the 2013 construction market outlook report, the liquidation of construction companies is expected to continue due to the ongoing residential construction recession. Bankruptcies of construction companies have a greater social impact compared to other industries. However, due to the different nature of the capital structure and debt-to-equity ratio, it is more difficult to forecast construction companies' bankruptcies than that of companies in other industries. The construction industry operates on greater leverage, with high debt-to-equity ratios, and project cash flow focused on the second half. The economic cycle greatly influences construction companies. Therefore, downturns tend to rapidly increase the bankruptcy rates of construction companies. High leverage, coupled with increased bankruptcy rates, could lead to greater burdens on banks providing loans to construction companies. Nevertheless, the bankruptcy prediction model concentrated mainly on financial institutions, with rare construction-specific studies. The bankruptcy prediction model based on corporate finance data has been studied for some time in various ways. However, the model is intended for all companies in general, and it may not be appropriate for forecasting bankruptcies of construction companies, who typically have high liquidity risks. The construction industry is capital-intensive, operates on long timelines with large-scale investment projects, and has comparatively longer payback periods than in other industries. With its unique capital structure, it can be difficult to apply a model used to judge the financial risk of companies in general to those in the construction industry. Diverse studies of bankruptcy forecasting models based on a company's financial statements have been conducted for many years. The subjects of the model, however, were general firms, and the models may not be proper for accurately forecasting companies with disproportionately large liquidity risks, such as construction companies. The construction industry is capital-intensive, requiring significant investments in long-term projects, therefore to realize returns from the investment. The unique capital structure means that the same criteria used for other industries cannot be applied to effectively evaluate financial risk for construction firms. Altman Z-score was first published in 1968, and is commonly used as a bankruptcy forecasting model. It forecasts the likelihood of a company going bankrupt by using a simple formula, classifying the results into three categories, and evaluating the corporate status as dangerous, moderate, or safe. When a company falls into the "dangerous" category, it has a high likelihood of bankruptcy within two years, while those in the "safe" category have a low likelihood of bankruptcy. For companies in the "moderate" category, it is difficult to forecast the risk. Many of the construction firm cases in this study fell in the "moderate" category, which made it difficult to forecast their risk. Along with the development of machine learning using computers, recent studies of corporate bankruptcy forecasting have used this technology. Pattern recognition, a representative application area in machine learning, is applied to forecasting corporate bankruptcy, with patterns analyzed based on a company's financial information, and then judged as to whether the pattern belongs to the bankruptcy risk group or the safe group. The representative machine learning models previously used in bankruptcy forecasting are Artificial Neural Networks, Adaptive Boosting (AdaBoost) and, the Support Vector Machine (SVM). There are also many hybrid studies combining these models. Existing studies using the traditional Z-Score technique or bankruptcy prediction using machine learning focus on companies in non-specific industries. Therefore, the industry-specific characteristics of companies are not considered. In this paper, we confirm that adaptive boosting (AdaBoost) is the most appropriate forecasting model for construction companies by based on company size. We classified construction companies into three groups - large, medium, and small based on the company's capital. We analyzed the predictive ability of AdaBoost for each group of companies. The experimental results showed that AdaBoost has more predictive ability than the other models, especially for the group of large companies with capital of more than 50 billion won.

• Applied Biological Chemistry
• /
• v.16 no.1
• /
• pp.18-30
• /
• 1973

The flag and lower leaves (4th or 5th) of rice plant from the field of NPK simple trial and from three low productive area were analyzed in order to find out certain diagnostic criteria of nutritional status at harvest. 1. Nutrient contents in the leaves from no fertilizer, minus nutrient and fertilizer plots revealed each criterion for induced deficiency (severe deficient case induced by other nutrients), deficiency (below the critical concentration), insufficiency (hidden hunger region), sufficiency (luxuary consumption stage) and excess (harmful or toxic level). 2. Nitrogen contents for the above five status was less than 1.0%, 1.0 to 1.2, 1.2 to 1.6, 1.6 to 1.9 and greater than 1.9, respectively. 3. It was less than 0.3%, 0.3 to 0.4, 0.4 to 0.55 and greater than 0.55 for phosphorus $(P_2O_5)$ but excess level was not clear. 4. It was below 0.5%, 0.5 to 0.9, 0.9 to 1.2, 1.2 to 1.4 and above 1.4 for potassium. 5. It was below 4%, 4 to 6, 6 to 11 and above 11 for silicate $(SiO_2)$ and no excess was appeared. 6. Potassium in flag leaf seemed to crow out nitrogen to ear resulting better growth of ear by the inhibition of overgrowth of flag leaf. 7. Phosphorus accelerated the transport of Mg, Si, Mn and K in this order from lower leaf to flag, and retarded that of Ca and N in this order at flowering while potassium accelerated in the order of Mn, and Ca, and retarded in the order of Mg, Si, P and N at milky stage. 8. Transport acceleration index (TAI) expressed as (F_2L_1-F_1L_2)\;100/F_1L_1$ where F and L stand for other nutrient cotents in flag and lower leaf and subscripts indicate the rate of a nutrient applied, appears to be suitable for the effect of the nutrient on the translocation of others. 9. The content of silicate $(SiO_2)$ in the flag was lower than that of lower leaf in the early season cultivation indicating hinderance in translocation or absorption. It was reverse in the normal season cultivation. 10. The infection rate of Helminthosporium frequently occurred in the potassium deficient field seemed to be related more to silicate and nitrogen content than potassium in the flag leaf. 11. Deficiency of a nutrient occured simultaniously with deficiency of a few other ones. 12. Nutritional disorder under the field condition seems mainly to be attributed to macronutrients and the role of micronutrient appears to be none or secondary.

• Journal of the Korean Institute of Traditional Landscape Architecture
• /
• v.38 no.3
• /
• pp.25-34
• /
• 2020

This study examined the definition and major values, the designated procedures and types, and the designation trend in Korean scenic sites and national parks in the United States. Based on this, the analysis of the characteristics of the designation of the two natural heritages. The results are as follows; First, Scenic Sites has characteristics of complex heritage that includes academic, historical, and humanities values on the basis of landscape. As a natural heritage based on public nature, the U.S. National Park aims to contribute to the people's natural heritage and satisfy both ecological and historical values through the protection of the landscape. Second, the designation of a scenic sites are decided through deliberation by the Cultural Heritage Committee after the request of the owner, manager, or local government or by the authority of the head of the Cultural Heritage Administration. The designated survey is divided into basic resource surveys and resource surveys by type. Since the initial designation of the Sogeumgang Mountain in Cheonghakdong, Myeongju in 1970, the number of designated scenic sites was low until the 2000s, but the number of designated scenic sites has increased rapidly since 2006 due to the policy to promote the scenic site, and the proportion of natural and historical and cultural scenic sites has been balanced. The designation of the U.S. national park is decided by the Congress or the president, and the National Park Service makes a series of decisions on whether to conduct a special resource study of provisional resources through a preliminary inspection survey, whether to satisfy the criteria for designation of national parks based on the results of special resource research, and to prioritize them. The U.S. National Parks have been expanded not only by Congress but also by the president's empowerment to designate them as national monuments. With the integrated operation of the National Park Service, the number of designated cases increased as the national park included the heritage sites under the control of various ministries. In addition, a number of historical areas were designated by the enactment of the Historical Site Act, and recreational areas were designated to provide leisure space and classified and managed in a total of 18 units. Third, the comparison of the designation characteristics of the two heritage properties confirmed that the designation of natural heritage with complex value, the classification of types according to complementary designation system and resource characteristics, the establishment of the competent ministry and the balancing of the heritage according to the designation policy. The two heritages had the characteristics of complex natural heritages that met ecological, historical and academic values at the same time based on landscape and public nature. In addition, both countries have identified a system for deliberating the designation of heritage through a basic resource survey and an in-depth designation survey, and classified each type according to the characteristics of the resource. In addition, the policies for promoting scenic sites in Korea and the integrated operation of the National Park Service in the U.S. influenced the designated aspects of the two heritage sites, balancing natural heritage with historical and cultural heritage. Fourth, the resource types and conservation management methods of Scenic site and National Park were largely related. The natural areas of the U.S. National Park include types of natural monuments in Korea as major resources, and have characteristics similar to natural scenic sites. In addition, historical resources were similar to the criteria for designation of historical and cultural scenic sites in terms of landscape, and the aspects of war and celebrity-related relics were related to the types of historic sites. In terms of conservation management, the natural area of the U.S. national park has a way of keeping the original ecosystem intact, but the Korean natural heritage protection system is likely to be useful for focusing on the resource of viscosity. Meanwhile, historical resources include historical sites and historical and cultural scenic sites in the traditional era, but historical relics in the U.S. National Parks have set a time limit to modern times for war history and celebrity-related relics, and the active provision of entertainment programs based on existing resources was derived as a difference.

• Journal of the Korean Society of Radiology
• /
• v.13 no.3
• /
• pp.371-379
• /
• 2019

The purpose of this study was to investigate the effect of the change in the X-ray condition on the entrance surface dose (ESD) and dose area product (DAP) in the cine imaging of coronary angiography (CAG), and to analyze the usefulness of the condition change on the dose relation and image quality by measuring and analyzing the Signal to Noise Radio (SNR) and Contrast to Nois Ratio (CNR) of the angiographic images taken by the Image J program. Data were collected from 33 patients (24 males and 9 females) who underwent CAG at this hospital from November 2017 to March 2018. In terms of imaging condition and data acquisition, the ESD and DAP of group A with a high tube current of 397.2 mA and group B with a low tube current of 370.7 mA were retrospectively obtained for comparison and analysis. For the SNR and CNR measurement and analysis via Image J, the result values were derived by substituting the obtained data into the formula. The correlations among ESD and DAP according to the change in the imaging condition, SNR, and CNR were analyzed by using the SPSS statistical analysis software. The relationships of groups A and B, having a difference in the imaging condition, mA, with ESD ($A:483.5{\pm}60.1$; $B: 464.4{\pm}39.9$) and DAP ($A:84.3{\pm}10.7$; $B:81.5{\pm}7$) were not statistically significant (p>0.05). In the relationships with SNR and CNR based on Image J, the SNR ($5.451{\pm}0.529$) and CNR ($0.411{\pm}0.0432$) of the images obtained via the left coronary artery (LCA) imaging of group B showed differences of $0.475{\pm}0.096$ and $-0.048{\pm}0.0$, respectively, from the SNR ($4.976{\pm}0.433$) and CNR ($0.459{\pm}0.0431$) of the LCA of group A. However, the differences were not statistically significant (p<0.05). In the SNR and CNR obtained via the right coronary artery (RCA) imaging, the SNR ($4.731{\pm}0.773$) and CNR ($0.354{\pm}0.083$) of group A showed increased values of $1.491{\pm}0.405$ and $0.188{\pm}0.005$, respectively, from the SNR ($3.24{\pm}0.368$) and CNR ($0.166{\pm}0.033$) of group B. Among these, CNR was statistically significant (p<0.05). In the correlation analysis, statistically significant results were shown in SNR (LCA) and CNR (LCA); SNR (RCA) and CNR (RCA); ESD and DAP; ESD and sec; DAP and CNR (RCA); and DAP and sec (p<0.05). As a result of the analyses on the image quality evaluation and usefulness of the dose change, the SNR and CNR were increased in the RCA images of the CAG obtained by increasing the mA. Based on the result that CNR showed a statistically significant difference, it is believed that the contrast in the image quality can be further improved by increasing the mA in RCA imaging.

• 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.