• 한국정보통신학회:학술대회논문집
• /
• 한국정보통신학회 2013년도 추계학술대회
• /
• pp.393-395
• /
• 2013

현대 사회에는 의학기술의 발전과 생활수준 향상 등으로 고령자들이 증가하고 있다. 고령자들의 낙상은 심한 경우 사망에 까지 이를 수 있는 상당히 큰 위협이 된다. 이러한 문제점을 해결하기 위해 낙상을 감지하는 여러 가지 알고리즘과 하드웨어 시스템의 필요성이 증가 하고 있으며 국내외에서 낙상 감지 시스템의 연구 결과가 발표 되고 있다. 본 논문에서는 3축 가속도 센서를 이용한 낙상 감지 시스템을 소개한다. 낙상 감지 시스템은 3축 가속도 센서 데이터로부터 몇 가지의 파라미터를 계산하여 낙상을 판별한다. 제안된 시스템을 이용하여 최대 98.3%의 sensitivity와 94.7%의 specificity 결과 값을 얻었다.

• 전기학회논문지
• /
• 제60권6호
• /
• pp.1204-1208
• /
• 2011

In this study, we proposed the system that calculates the algorithm with an accelerometer signal and detects the fall shock and it's direction. In order to gather the activity patterns of fall status and attach on the subject's body without consciousness, the device needs to be small. With this aim, it is attached on the right side of subject's waist. With roll and pitch angle which represent the activity of upper body, the fall situation is determined and classified into the posture pattern. The impact is calculated by the vector magnitude of accelerometer signal. And in the case of the elderly keep the same posture after fall, it can distinguish the situation whether they can stand by themselves or not. Our experimental results showed that 95% successful detection rate of fall activity with 10 subjects. For further improvement of our system, it is necessary to include tasks-oriented classifying algorithm to diverse fall conditions.

• 한국정보통신학회:학술대회논문집
• /
• 한국정보통신학회 2021년도 추계학술대회
• /
• pp.391-393
• /
• 2021

본 논문에서는 일상생활에서의 고령자에게 나타날 수 있는 낙상상황을 감지할 수 있는 텐서플로우를 이용한 장단기 메모리 기반 낙상감지 시스템에 대하여 소개한다. 낙상감지를 위해서 3축 가속도 센서 데이터를 이용하고, 이를 처리하여 다양한 파라미터화하며 일상생활 패턴 4가지, 낙상상황 패턴 3가지로 분류한다. 파라미터화한 데이터는 정규화 과정을 따르며, 학습이 진행된다. 학습은 Loss값이 0.5 이하가 될 때까지 진행된다. 각각의 파라미터인 θ, SVM (Sum Vector Magnitude), GSVM (gravity-weight SVM)에 대하여 결과를 산출한다. 가장 좋은 결과는 GSVM으로 Sensitivity 98.75%, Specificity 99.68%, Accuracy 99.28%로 가장 좋은 결과를 보였다.

• 한국정보통신학회논문지
• /
• 제25권11호
• /
• pp.1649-1654
• /
• 2021

본 논문에서는 고령자의 낙상상황을 감지할 수 있는 텐서플로우 장단기 메모리 기반 낙상감지 시스템의 정규화에 대하여 소개한다. 낙상감지는 고령자의 몸에 부착한 3축 가속도 센서 데이터를 사용하며, 총 7가지의 행동 패턴들에 대하여 학습하며, 각각 4가지는 일상생활에서 일어나는 패턴이고, 나머지 3가지는 낙상에 대한 패턴이다. 학습시에는 손실함수(loss function)를 효과적으로 줄이기 위하여 정규화 과정을 진행하며, 정규화 과정은 데이터에 대하여 최대최소 정규화, 손실함수에 대하여 L2 정규화 과정을 진행한다. 3축 가속도 센서를 이용하여 구한 다양한 파라미터에 대하여 정규화 과정의 최적의 조건을 제시한다. 낙상 검출율면에서 SVM을 이용하고 정규화 127과 정규화율 λ 0.00015일 때 Sensitivity 98.4%, Specificity 94.8%, Accuracy 96.9%로 가장 좋은 모습을 보였다.

• 로봇학회논문지
• /
• 제13권1호
• /
• pp.45-54
• /
• 2018

This paper presents a vision-based fall detection system to automatically monitor and detect people's fall accidents, particularly those of elderly people or patients. For video analysis, the system should be able to extract both spatial and temporal features so that the model captures appearance and motion information simultaneously. Our approach is based on 3-dimensional convolutional neural networks, which can learn spatiotemporal features. In addition, we adopts a thermal camera in order to handle several issues regarding usability, day and night surveillance and privacy concerns. We design a pan-tilt camera with two actuators to extend the range of view. Performance is evaluated on our thermal dataset: TCL Fall Detection Dataset. The proposed model achieves 90.2% average clip accuracy which is better than other approaches.

• 스마트미디어저널
• /
• 제7권4호
• /
• pp.79-89
• /
• 2018

In recent years, falls among elderly people have gained serious attention as a major cause of injuries. Falls often lead to fatal consequences due to lack of prompt response and rescue. Therefore, a more accurate fall detection system and an effective feature extraction technique are required to prevent and reduce the risk of such incidents. In this paper, we proposed an efficient feature extraction technique based on multiple sliding windows and validated it through a series of experiments using supervised learning algorithms. The experiments were conducted using the public datasets obtained from tri-axial accelerometers. The results depicted that extraction of the feature from adjacent sliding windows led to high accuracy in supervised machine learning-based fall detection. Also, the experiments conducted in this study suggested that the best accuracy can be achieved by keeping the window size as small as 2 seconds. With the kNN classifier and dataset from wearable sensors, the experiments achieved accuracy rates of 94%.

• 인터넷정보학회논문지
• /
• 제18권3호
• /
• pp.11-17
• /
• 2017

To the elderly, sudden falls are one of the most frightening accidents. If an accident occurs, a prompt action has to be taken to deal with the situation. Recently, there have been a number of attempts to detect sudden falls using acceleration sensors embedded in the mobile devices, such as smart phones and wrist-bands. However, using the sensor readings only, the detection rate of the falls is around 65%. Ordinary daily activities such as running or jumping could not be well distinguished from the falls. In this paper, we describe our attempts on improving the fall-detection rate. We implemented a wrist-band fall detection module, using a three-axis acceleration sensor. With the pattern matching on the fall signal-strength frequency spectrum, in addition to the conventional signal strength measurement, we could improve the detection rate by 9% point. Furthermore, by applying two wrist-bands in the experiment, we could further improve the detection rate to 82%.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제18권9호
• /
• pp.2782-2804
• /
• 2024

The integration of artificial intelligence technology with medicine has rapidly evolved, with increasing demands for quality of life. However, falls remain a significant risk leading to severe injuries and fatalities, especially among the elderly. Therefore, the development and application of computer vision-based fall detection technologies have become increasingly important. In this paper, firstly, the keypoint detection algorithm ViTPose++ is used to obtain the coordinates of human body keypoints from the camera images. Human skeletal feature maps are generated from this keypoint coordinate information. Meanwhile, human dense feature maps are produced based on the DensePose algorithm. Then, these two types of feature maps are confused as dual-channel inputs for the model. The convolutional gated recurrent unit is introduced to extract the frame-to-frame relevance in the process of falling. To further integrate features across three dimensions (spatio-temporal-channel), a dual-channel fall detection algorithm based on video streams is proposed by combining the Convolutional Block Attention Module (CBAM) with the ConvGRU. Finally, experiments on the public UR Fall Detection Dataset demonstrate that the improved ConvGRU-CBAM achieves an F1 score of 92.86% and an AUC of 95.34%.

• 전기전자학회논문지
• /
• 제18권1호
• /
• pp.134-139
• /
• 2014

노인의 생활안전 사고 유형 중 가장 높은 비율을 차지하는 낙상은 57.2%이상이 가정에서 발생하는 것으로 조사되었다. 본 연구에서는 실내 바닥의 진동을 측정, 분석하여 낙상의 유무를 판별하고자 하였으며, 이를 위해 압전필름과 연산증폭기로 증폭 및 필터링 회로를 제작하여 진동 센서 모듈을 구성하였다. 진동 센서 모듈에서 증폭 및 필터링 과정을 거친 진동 신호는 데이터 수집 장치를 통해 디지털 신호로 변환되어 PC로 전송된다. 진동 신호는 k-NN 분류기를 이용하여 낙상 유무를 판별한다. 피험자 10명을 대상으로 낙상 실험결과, 분류기는 93.6%의 인식율을 나타내었다. 제작된 센서 모듈은 낙상 검출에 유용한 것으로 판단된다.

• Journal of information and communication convergence engineering
• /
• 제22권2호
• /
• pp.139-144
• /
• 2024

In this study, four types of fall detection systems - designed with YOLOPose, principal component analysis (PCA), convolutional neural network (CNN), and long short-term memory (LSTM) architectures - were developed and compared in the detection of everyday falls. The experimental dataset encompassed seven types of activities: walking, lying, jumping, jumping in activities of daily living, falling backward, falling forward, and falling sideways. Keypoints extracted from YOLOPose were entered into the following architectures: RAW-LSTM, PCA-LSTM, RAW-PCA-LSTM, and PCA-CNN-LSTM. For the PCA architectures, the reduced input size stemming from a dimensionality reduction enhanced the operational efficiency in terms of computational time and memory at the cost of decreased accuracy. In contrast, the addition of a CNN resulted in higher complexity and lower accuracy. The RAW-LSTM architecture, which did not include either PCA or CNN, had the least number of parameters, which resulted in the best computational time and memory while also achieving the highest accuracy.