Fall Detection Depth-Based Using Tilt Angle and Shape Deformation

Fall Detection Depth-Based Using Tilt Angle and Shape Deformation

Fairouz Merrouche (University of Science and Technology USTHB, Algeria) and Nadia Baha (University of Science and Technology USTHB, Algeria)
DOI: 10.4018/978-1-7998-1204-3.ch039
OnDemand PDF Download:
No Current Special Offers


The population of elderly people is in growth. Falls risk their life, to disabilities, and to fears. Automatic fall detection systems provide them secure living; helping them to be independent at home. Computer vision offers efficient systems over many developed systems. In this article, the authors propose a new vision-based fall detection using depth camera. It combines human shape analysis, centroid detection and motion where it exploits the 3D information provided by a Kinect to compute the tilt angle to discriminate falls. Experimental tests were done with SDUFall dataset that contains 20 subjects performing five daily activities and falls, demonstrate the efficiency of the proposed system, and show that our method is promising achieving satisfactory results up to 84.66%.
Chapter Preview

In the last few years, various approaches have been put forward by many researches to improve fall detection solutions. Fall detection methods are subdivided into two classes: computer vision-based methods and non-computer vision method.

In this section, we will review some works proposed for fall detection starting by non-vision category. Many sensors have been exploited. In the work of (Cheng et al., 2013) the authors have attached two accelerometers to chest and thigh of 10 persons. Decision tree was applied to recognize posture transitions with threshold to detect falls. The main disadvantage of this method is to wear the sensor.

Another work was proposed in (Sannino et al. 2015) where a wearable accelerometer and a mobile device were used to monitor the patients in real time, the data recorded by the accelerometer is sent to the mobile device to analyze them. This approach has three phases. First, gather data and make a set of simulated falls with the corresponding class “fall” or “non-fall”. Second, knowledge extraction as a set of IF-THEN rules and the last phase consists in patients monitoring in the decisional layer in real time but this approach still preliminary, the tests were done for just three people.

In the work of (Alwan et al., 2006), a special piezoelectric sensor coupled to the floor is used, a binary fall signal can be generated in case of fall, and however, the main drawbacks can be false alarms.

In (Yazar et al., 2014) fall was detected based on three sensors consisting of a vibration sensor and two PIR sensors. The vibrations were converted into electrical signals and then Feature vectors from the vibration waveforms are extracted using the complex wavelet transform (CWT) and they are classified using support vector machine (SVM). PIR sensors were used as additional sensors to detect the infrared radiation emitted by moving objects in the room and that to eliminate falls alarms generated when the vibration sensor signal energy is concentrated in low-frequency bands such as vibration from door slams and other similar events.

In recent work in (Jokanovic et al., 2016) a new method was proposed based on radar technologies to detect falls, the authors proposed to use time-frequency (TF) which is able to reveal higher order velocity for different part of human body and based on deep learning which learns and captures the properties of the TF signatures without human intervention then it is fed to a classifier achieving 87% success rate.

Complete Chapter List

Search this Book: