Motion sensors for activity recognition in an ambient-intelligence scenario

Abstract

In recent years, Ambient Intelligence (AmI) has attracted a number of researchers due to the widespread diffusion of unobtrusive sensing devices. The availability of such a great amount of acquired data has driven the interest of the scientific community in producing novel methods for combining raw measurements in order to understand what is happening in the monitored scenario. Moreover, due the primary role of the end user, an additional requirement of any AmI system is to maintain a high level of pervasiveness. In this paper we propose a method for recognizing human activities by means of a time of flight (ToF) depth and RGB camera device, namely Microsoft Kinect. The proposed approach is based on the estimation of some relevant joints of the human body by using Kinect depth information. The most significative configurations of joints positions are combined by a clustering approach and classified by means of a multi-class Support Vector Machine. Then, Hidden Markov Models (HMMs) are applied to model each activity as a sequence of known postures. The proposed solution has been tested on a public dataset while considering four different configurations corresponding to some state-of-the-art approaches and results are very promising. Moreover, in order to maintain a high level of pervasiveness, we implemented a real prototype by connecting Kinect sensor to a miniature computer capable of real-time processing

Human Activity Recognition Process Using 3-D Posture Data

“…Following these considerations, our perspective is to consider the Kinect as the primary sensor to transparently gather observations about users' behavior [8].…”

“…The main limitations of this approach are the adoption of a complex model for representing the joints and the consequent need for reducing the dimensionality of the feature vectors by means of LDA. In [8], we observed that if the feature space already contains an optimal set of features, the attempt of further reducing such a space by means of principal component analysis or LDA does not increase the overall performance of the system, but may instead prevent the achievement of real-time processing.…”

“…In [8], due to the sensitiveness of the IR sensor, some joints are misdetected if two segments overlap (e.g., hands touching other body parts), or not detected at all due to the presence of objects between the sensor and the user. For this reason, we evaluated the system by measuring the recognition rates achieved on a limited number of selected subset of joints.…”

Simultaneous segmentation and classification of human actions in video streams using deeply optimized Hough transform

“…However there are three problems. First, even if [38] presents a segmentation algorithm, algorithm TUM UTKAD HMM [48] 90.9 NN [11] 91.5 SVM [9] 90.4 sparse SVM [41] 90.9 ISM 65.5 64.1 unscaled SVM 80.6 79.8 DOHT 88.5 82.4 Table 9: Results on already segmented data on several datasets.…”

Online social networks as sensors in smart environments

“…Ambient Intelligence and Smart Environments are impulsed by ubiquitous computing and take advantage of the ease of collecting data from numerous devices in order to produce tasks such as optimisation of energy consumption [1], [2], recognition of human activity and preferences [3], [4], aid the elderly or persons with health problems [5], or even increase the lifestyle of blind people [6]. Although this type of implementations require physical sensors to capture data, there is another way to gather data without the need of a physical sensor.…”