Recognition of Day Night Activity Using Accelerometer Principals with Aurdino Development Board Swapna Shivaji Arote R.S. Bhosale S.E. Pawar Dept. of Information Technology, Dept. of Information Technology, Dept. of Information Technology, Amrutvahini college of Engg., Amrutvahini college of Engg., Amrutvahini college of Engg., Sangamner (M.S) India-431005 Sangamner (M.S) India-431005 Sangamner (M.S) India-431005 Email: aroteswapna@gmail.com Email: bhos_raj@rediffmail.com Email: pawar.suvarna@gmail.com Abstract: We often hear that senior citizens, disabled peoples, patients and kids falling when they are out. It s difficult to monitor them continuously; we restrict them to go out without they are being accompanied. We identified this problem and thought of automatic system which will keep monitoring there real time activities and can send you the alerts when there guardians are actually seating back in the room doing their work and in case of emergency this alert can get them immediate help which will surely help them to reduce casualty. This can give freedom to this older people or kids to go out on their own. An accelerometer accompanied with GSM instant messaging module which is real time, portable, and wearable, low-cost and with high accuracy rate. This device can monitor the movements and send alerts in case of emergency. Keywords: fall detection, Kinect, older adults, Tri-axial accelerometer I. INTRODUCTION Falls are the most commonly happening accident among the elder people, disables and patients. It often impacts to the sufferers and people around them. Such situations incur high medical costs. People, who have earlier met this kind of accident often fear a lot and wish not to have a new fall and this fear of another fall take them into idleness or a kind of social isolation, many of the times they wish to go out but can t really go just because of the fear of collapse, We thought of this savior problem and solution which can detect such kind of accidents and generate alert which can help them get immediate assistance. Such falls can be detected by 2 ways: First method is Visual observation with some visual device which can observe the person continuously and then image analyzer which can analyze the fall of the person and generate alarm. But this kind of system can have some flaws like it will capture all the personal activities. The scope of this kind of device is limited to monitoring region (Majorly indoors) cannot be carried all the time with the person Information provided by device is quite accurate but such equipment will cost higher. w days, Microsoft launched its product the Kinect sensors [5] that include an IRFF projector and camera, thus this system functionally work as a vision-based sensor under all lighting conditions. It is useful to gathering activity data during all times of the day. Microsoft kinect camera is enabling to gather information during night times. This is biggest disadvantage of this system. Second method is through sensor to overcome the flaws of visual monitoring equipment we thought of devising electronic sensor which will important to detect fall and send out alert to call for help. But if a fallen person is unconscious and unable to call for help it can lead to permanent injure and even death. Therefore there is a need of self-directed fall detectors that are capable of triggering an alarm automatically without any interference of the victim and transferring this information to a remote site. So that9 the fallen person should get immediate medical help. Based on these problems, the main purpose, of the proposed system is to develop a small, comfortable, and user friendly device, as well as an automatic fall detector system that will help older people to handle this problem. In this paper the fall detector is a real time working model for detecting the fall, using accelerometer. It will be capable of automatically detecting a variety of body falls and sending an alarm to a remote terminal. The user can manually generate the alarm in case of necessity or, inversely, he or she can cancel an automatically generated alarm in case of false fall detection. It will consist of GSM Technology attached with this system for sending the real time status to the respective places (Family Doctor, Relatives etc.) As for fall detection through an accelerometer, the scope of activity is relatively unrestricted; the device is easily attached to the human body, and the cost of equipment is low. This paper is organized as follows. Section II discusses related work about fall detection mechanism. Section III presents the proposed system of fall detection. Section IV discusses result of the proposed scheme. Section V presents the conclusion this study. II. RELATED WORK Accelerometer [1] has been used in various studies and applications to objectively monitor a range of human movement, for example to measure metabolic energy expenditure, physical activity levels, balance and postural sway, gait, and to detect falls.
According to the survey done in many countries, more and more elderly people will meet in accident of falls at home and they will require emergency help. The main problem arise when older person living alone at home. To resolve this problem for safety purpose, author proposes a new method to detect falls at home, using a multiple-cameras [9] network in home for reconstructing the 3-D shape of person. Falls are found by analyzing the volume distribution along the vertical axis, and an alarm is triggered when the distribution is not normally that is near to the floor during a predefined time period, which shows that a person has fallen on the ground. For this purpose, author proposed a method that is capable of dealing with several occlusions that could occur in personal houses.this method of detecting fall is based on two ideas. First, they use the occlusion-resistant algorithm to detect the fall if a person is lying on the ground or not. Second, they defined the original and simple idea of vertical volume distribution ratio [2]. Dividing the volume that is below a given height by the total volume from this ratio is obtained. For activity monitoring person through video for eldercare deal with the some issues. First, activity monitoring should be unobtrusive and privacyprotecting. For unobtrusiveness, the size and total number of cameras need to be as small as possible so that they can be easily and hidden in the living environment. In this paper, author installs a single small fisheye [3] camera in the living room. For privacy protection, they develop a fast and efficient silhouette extraction algorithm to extract and block-out the person in the video frame. In this paper author developed an accurate and robust silhouette extraction and human tracking algorithm which is used to effectively remove shadow and handle dynamic background changes in an indoor living environment. System for capturing regular, in-home step measurements using an environmentally mounted depth camera, the Microsoft Kinect, is presented. In This paper a single Kinect sensor and computer were placed in the apartments of older adults in an independent living. For the purpose of continuous monitoring. In this paper have investigated algorithms for measuring gait parameters using the Kinect sensor without the use of skeletal tracking, validating the measurements against a Vicon markerbased motion [4] capture system in a laboratory setting, and extending this approach to capture gait in realworld, dynamic environments, specifically the homes of older adults. III. PROPOSED FALL DETECTION METHODS The proposed monitoring system show in figure1. consists of a hardware unit attachable to the waist of the user and a computational-logic microcontroller for classifying user's movements and detecting any possible falls. In case of the falls, the system also sends out an alarm to the appropriate response unit. The first part of the system includes a fall-detecting band for extracting and processing signals obtained from the triaxial accelerometer and the tilt sensors. According the block diagram, an accelerometer ADXL335 will be used for detection the fall in X, Y and Z axis. Here these sensors will send the analog signal to the microcontroller for its logical manipulation for detecting the real time status of the body to the home server to update the display information. The device also includes an emergency help button display the fall alert and emergency signal. The second part of the system consists of a GSM modem system that will be attached with the microcontroller for sending the message to the respective places along with the location of user. The server first received a fall alert, and then it generates alarm Accelerometer Microcontroller SIM 900A CC2500 Wireless Sensor Network CC2500 Micro - controller Figure 1 : Block diagram of the proposed fall detection system. Fall detection technique. The extracted acceleration values by the tri-axial accelerometer will be input to microcontroller for computing its signal vector magnitude. The three axes i.e. x, y, z will each produce a different acceleration value, based on these value use stage analysis to evaluate users movement Processor Modem USB
signals. The main distinctiveness of a wireless sensor network perfectly well suitable for a fall detection system based on the wearable approach. The silhouette, volume and load of the nodes allow them to be worn easily by older person. Movements will be classified into normal movements and abnormal movements. According to acceleration signals, normal movements will be continuous and cadent movement signals, while abnormal movements shall be recognized as fall signals. In this method check the value of three axes with respective defined threshold. If value is below the threshold then generate alarm. Each device is generally battery-powered and can be instrumented with one or more sensors which make possible of gaining physical data. If(Z<threshold) { Another event within 3 Sec Start Movement DigitalWrite(ledPin, Low); } Else { Calculate values from accelerometer before the latest event DigitalWrite(ledPin,High); } During a fall acceleration pattern is a decrease in acceleration threshold followed by an increase acceleration threshold. This is happen because an accelerometer at during free fall 0 g and rest registers 1 g (the Earth s gravity). When a person starts falling, the acceleration decreases from 1 g to around 0.5 g. The highest and lowest acceleration within a 3 second measured. If the difference the difference between the highest and the lowest exceed 1 g and the highest come after the lowest, then fall is detected. For this purpose step sequence algorithm is used. In this algorithm coordinate values are taken from the accelerometer to define conditions. After defining conditions algorithm check the values of coordinate step by step. When a older person falls, there is sudden a large change in position from before fall & after. It is something like would be from standing upright to lying flat, It is right angle change. Calculate values of accelerometer axes at last event Change in Axes Generate Alert End Figure 2: Flow chart of Fall detection System
Object can be monitor behind the wall unlike Kinect system. The main reason for not using the Kinect SDK is the limited range of the skeletal tracking, approximately 1.5 to 4 meters from the Kinect. It s ability to determine different activities under different environments. Speed up the activity state identification process and lead to faster fall detection. IV. RESULT TABLE I : Analysis chart for the accelerometer coordinate activities of straight, bend, flat CONDITIONS AXES X Y Z Straight 272 296 332 Bend 308 315 396 CONCLUSION There is various fall detection systems detect effectively, but not Suitable for real time applications because of the larger number of sensor nodes attached to human body and uncomfortable wearing condition. The proposed system will be a portable device mounted on the waist of user, having sensors consisting of accelerometer. The propose fall detection system can be regarded as alternative device to the existing detection approaches, since the device provides the comfortable wearing, is less complex as compared to other devices, fast fall response and will be more accurate and economical. These systems overcome the limitation of limited range of object tracking. REFERENCES [1] M.J. Mathie, A.C. Coster, N.H. Lovell, and B.G. Celler, Accelerometry: providing an integrated, practical method for long-term, ambulatory monitoring of human movement, Physiol. Meas. 2004 Apr VOL-25(2):R1-20 Flat 332 336 404 400 300 200 X 100 Y 0 X Z Z [2] E. Auvinet et al., "Fall detection with multiple cameras An occlusion resistant method based on 3-d silhouette vertical distribution," IEEE Trans. on Info. Tech. in Biomedicine, 15(2):290-300, VOL-12; 2011. [3] Z. Zhou, X. Chen, Y. C. Chung, Z. He, T. X. Han, and J. M. Keller, Activity analysis, summarization, and visualization for indoor human activity monitoring, IEEE Trans. Circuits Syst. Video Technol., vol. 18, no. 11, pp. 1489 1498, v. 2008. [4] E. Stone and M. Skubic, Unobtrusive, Continuous, In-Home Gait Measurement Using the Microsoft Kinect, IEEE Transactions Biomedical Engineering, vol-60(10):2925-2932,,2013. [5] T. Banerjee, J. M. Keller, Z. Zhou, M. Skubic, and E. Stone, Day or Night Activity Recognition From Video Using Fuzzy Clustering Techniques IEEE transactions on fuzzy systems, VOL. NO. 3, JUNE 2011. Figure 3: Comparison of straight, bend, flat conditions Figure 3 shows the comparison between the straight, flat and bend conditions. The abnormal activities like sleeping, sitting etc are detected using the particular threshold values given. The alarm is set in the form of message or voice alarm. This SMS can be sent to doctor or care taker. If x is less 200 & greater than 300 then person, y is greater and less than 400 and z value less than or greater than 400 then tilt angle is 40, 60, 80 degree. If Tilt angle value is between 30 to 40 degree and alarm is generated i.e. fall is detected. [6] D. E. Gustafson and W. C. Kessel, Fuzzy clustering with a fuzzy covariance matrix, in Proc. IEEE Annu. Conf. Decision Control, San Diego, CA, USA, VOL-3 1979, pp. 761 766. [7] M. J. Lesot and R. Kruse, Gustafson-Kessel-like clustering algorithm based on typicality degrees, presented at the Int. Conf. Inf. Process. Manag. Uncertainty, Paris, France, 2006. [8] Iosifidis,A. Tefas, and I. Pitas, Eating and drinking activity recognition based on discriminant analysis of fuzzy distances and activity volumes, in Proc. IEEE Int. Conf. Acoust., Speech Signal Process., Mar. 25 30, 2012, pp. 2201 2204.
[9] M.Goffredo, M. Schmid, S. Conforto,M. Carli, A. Neri, and T. D Alessio, Markerless human motion analysis ingauss- Laguerre transform domain:an application to sit-to-stand in young and elderly people, IEEE Trans. Inf. Technol. Biomed., vol. 13, no. 2, pp. 207 216, Mar. 2009. [10] T. Banerjee, J. M. Keller, M. Skubic, and C. C. Abbott, Sitto-stand detection using fuzzy clustering techniques, in Proc. IEEE Int. Conf. Fuzzy Syst./World Conf. Comput. Intell., 2010, pp. 1 8. [11] T. Banerjee, J. M. Keller, Z. Zhou, M. Skubic, and E. Stone, Activity Segmentation of infrared images using fuzzy clustering techniques, presented at the World Conf. Soft Comput., San Francisco, CA, USA, May 2011 [12] A. El Maadi and X. Maldague, Outdoor infrared video surveillance: A novel dynamic technique for the subtraction of a changing background of IR images, Infrared Phys. Technol., vol. 49, no. 3, pp. 261 265, 2007.