1. Progress Seminar
권순빈

2. 연구 진행 상황 보고서 2주전 계획 연구 결과 문제점 및 대책 목표 및 계획 CPF 정형외과 낙상 혈당센서
논문 제출 (JMIR mHealth and uHealth)
논문 제출
Response to Reviewer’s comment
문제점 및 대책
논문 reject
목표 및 계획

3. 낙상 Reviewer 1 (Major revision / Resubmit encouraged)
Encourage the author to add the results (quantitatively) of method in abstract
The hardware in Fig.1 should be expanded with a block diagram to introduce the system frame.
Data processing flow chart is encouraged to be added, and which makes the algorithm is better to be understood.
The English should be further polished, some sentences should be written in a better format.
The figures & tables should be improved.

4. 낙상
The hardware in Fig.1 should be expanded with a block diagram to introduce the system frame.
The hardware in Fig.1 should be expanded with a block diagram to introduce the system frame.

5. 낙상 Reviewer 2 (Minor revision / Accept)
Add some clearly formulated hypotheses from the literature review to set up the conducted experiments
Add more recent literatures (2개를 추천해주었음) + add literature summarizing table
Add dimension information about the hardware, explain how each features were extracted in more detail, justify why only used 3 ML, briefly explain performance indicator.
Delete table I and add raw signal of each fall
Encourages the authors to either compare or contract the results to previous studies.
Justifiable reasons to support why your methodology could be applied to the energy-efficient wearable device
Add limitation & provide future directions

6. 낙상
Add some clearly formulated hypotheses from the literature review to set up the conducted experiments
Original
Revised
Materials and Methods Section B
The fall experiment was designed based on the five most common types of fall with respect to elderly adults in daily life, namely incorrect weight shifting, trip or stumble, hit or bump, loss of support, and collapse [26]. Each fall type was performed nine times for each subject with three different falling speeds, fast, medium, and slow, for each type of fall.
The fall experiment was designed based on the five most common types of fall with respect to elderly adults in daily life, namely incorrect weight shifting, trip or stumble, hit or bump, loss of support, and collapse [26]. Robinovitch et al. have reported these fall types by observing 227 falls from 130 individuals at the long-term care facilities. The participants from this study watched the supplementary video of the previous studied for few times. After watching the video, the participants practiced each fall type for several times to get familiar with the protocol. Each fall type was performed nine times for each subject with three different falling speeds, fast, medium, and slow, for each type of fall.

7. 낙상
Add more recent literatures (2개를 추천해주었음) + add literature summarizing table
Author
Year
Sensor Type
Feature
Algorithm
Performance
Miaou et al
2006
Omni-camera
The change in ratio of height and width from the silhouettes
Threshold algorithm
Accuracy: 81%
Bian et al
2015
Depth-camera
3-D trajectory of the head joint
Support vector machine
Accuracy: 98%
Rougier et al
2011
1. Human centroid height relative to the ground 2. Body velocity
Occlusion robust method
Accuracy: 99%
Yu et al.
2013
Traditional camera
Extracted ellipses and shape-structure feature from the image
one-class support vector machine
Ture positive rate: 100 False negative rate: 3%
Kangas et al
2008
Three triaxial accelerometers
1. Total sum vector 2. Dynamic sum vector 3. Vertical accereration 4. Falling index
Lai et al
Six triaxial accelerometers
1. Sum vector magnitude 2. Activity single magnitude Area
92.92%
Aziz et al
four triaxial accelerometers
Mean and variance in X, Y, and Z acceleration traces
Linear discriminant analysis algorithm
Sensitivity: 83% Specificity: 89%
Yang et al
2017
IMU sensor
1. Stride time 2. Stride distance 3. Average velocity 4. Maximum foot clearance 5. Stance ratio 6. Swing ratio
Gait abnormality score NA
Mao et al
1. RMS of acceleration 2. Orientation of the user
Threshold detection
Accuracy: 100%
Nari et al
2016
1. Single vector magnitude 2. Angular velocity
Sensitivity: 90% Specificity: 87%
Antwi_Afari et al
2018
Insole pressure system
1. Mean pressure 2. Peak pressure 3. Pressure-time integral 4. A/P COP 5. M/L COP
Observed statistical significance of the five features

8. 낙상 Delete table I and add raw signal of each fall
Fig. 3. Description of how each feature was extracted from the IMU data.
Fig. 3. Description of how each feature was extracted from the IMU data. (A) shows angular velocity in X and Y direction. Feature 1 through 10 were calculated by subtracting first and last value, first and maximum value, first and minimum value, last and maximum value, and last and minimum value of each X and Y direction. (B) show the acceleration data in X, Y, and Z direction. Feature 11 was determined based on the order of last value as shown in Eq. (8). Feature 12 was determined based on the order of last value as shown in Eq. (9)