This paper proposes a system for the early automatic recognition of health problems that manifest themselves in distinctive form of gait. Purpose of the system is to prolong the autonomous living of the elderly at home. When the system identifies a health problem, it automatically notifies a physician and provides an explanation of the automatic diagnosis. The gait of the elderly user is captured using a motion-capture system, which consists of body-worn tags and wall-mounted sensors. The positions of the tags are acquired by the sensors and the resulting time series of position coordinates are analyzed with machine-learning algorithms in order to recognize a specific health problem. Novel semantic features based on medical knowledge for training a machine-learning classifier are proposed in this paper. The classifier classifies the user’s gait into: i) normal, ii) with hemiplegia, iii) with Parkinson’s disease, iv) with pain in the back and v) with pain in the leg. The studies of i) the feasibility of automatic recognition and ii) the impact of tag placement and noise level on the accuracy of the recognition of health problems are presented. The experimental results of the first study (12 tags, no noise) showed that the k-nearest neighbors and neural network algorithms achieved classification accuracies of 100%. The experimental results of the second study showed that classification accuracy of over 99% is achievable using several machine-learning algorithms and 8 or more tags with 0-15 mm standard deviation of noise. The results show that the proposed approach is accurate and can be used as a guide for further studies in the increasingly important area of Ambient Assisted Living. Since the system uses semantic features and an artificial-intelligence approach to interpret the health state, provides a natural explanation of the hypothesis and is embedded in the domestic environment of the elderly person; it is an example of the semantic ambient media for Ambient Assisted Living.

The term ambient media was in its beginning used only for ambient advertising. Nowadays it denotes the media environment and the communication of information in ubiquitous and pervasive environments. With the addition of intelligence, the new field of semantic ambient media was established. In recent years, the field of semantic ambient media has spread its span from only a few sub-areas, such as ambient advertising, to new ones, such as ambient-assisted living (AAL) and health-monitoring media, significantly supported by intelligence. The study presented in this paper provides an advanced introduction to the field of semantic ambient media including the solutions for threat issues and illustration of success stories of the field. It conducts a survey of the related work and presents a thorough discussion of it. The related work is grouped according to the coverage of the principles of semantic ambient media. Based on the state-of-the-art research, the future possibilities of the field are demonstrated, especially for the ambient-assisted living, audio-visual rendering of media objects, user design principles and the society impact of the field. The paper provides ideas for impacting ambient media and directions and questions for further research. It also discusses the potential of the combination of several research studies.

Background: The number of elderly people in need of help with the activities of daily living in the EU is rapidly increasing, while the number of young workers is decreasing. Elderly care will, therefore, also have to be provided by intelligent computer systems.

Methods: A prototype elderly-care system, developed at the Jožef Stefan Institute, mostly as a part of the Confidence project, is presented. The prototype detects falls and behavior changes in the elderly. It learns from experience and is based on intelligent interpretation of movement patterns. Three sets of tests were performed to evaluate its properties on various subjects when engaged in normal activities, falling and imitations of several health problems under medical supervision. The key novelty was in location-based sensors and advanced intelligent methods.

Results: The prototype using the Ubisense sensor system, which detects the locations of tags worn on the body, correctly recognized 96% of falls, significantly outperforming simple accelerometer-based systems. In addition, it recognized up to 99% of abnormal behavior.

Conclusions: Experimental results showed that an intelligent system coupled with advanced location sensors can achieve the level of performance needed in real life. The system offers significantly better performance than commercially available solutions, and once the price of sensors decreases, its widespread application seems likely.

Key words: fall recognition, 24-hour caregiver system, ambient assisted living, artificial intelligence, public healthcare

In Europe, in particular, growing numbers of elderly need sustainable elderly care, which the young are not able to provide. As an alternative, elderly care can be provided through home-based automatic health-monitoring systems. Here we propose data-mining algorithms in a system for the automatic recognition of health problems through the analysis of gait. The gait of the elderly is captured using a motion-capture system and the resulting time series of position coordinates are analyzed with a data-mining approach in order to classify the captured gait into: 1) normal, 2) with hemiplegia, 3) with Parkinson’s disease, 4) with pain in the back and 5) with pain in the leg. We propose and analyze four data-mining approaches: 1) CML – Classic machine-learning approach with raw sensor data, 2) SCML – Classic machine-learning approach with semantic attributes, 3) MDTW – Multidimensional dynamic time-warping approach with raw sensor data and 4) SMDTW – Multidimensional dynamic time-warping approach with semantic attributes.

Semantic Ambient Media are the novel trend in the world of multimedia and will likely shape the upcoming years in terms of media consumption, interaction, and modeling smart environments. This work reviews the state of the art of the Semantic Ambient Media field and analyzes ambient media by considering business models, content and media, interaction design and consumer experience, and methods and techniques that are important to create this new form of media. The study also creates a vision for the Semantic Ambient Media in the near future.

Due to the rapid aging of the European population, an effort needs to be made to ensure that the elderly can live longer independently with minimal support of the working-age population. The Confidence project aims to do this by unobtrusively monitoring their activity to recognize falls and other health problems. This is achieved by equipping the user with radio tags, from which the locations of body parts are determined, thus enabling posture and movement reconstruction. In the paper we first give a general overview of the research on fall detection and activity recognition. We proceed to describe the machine learning approach to activity recognition to be used in the Confidence project. In this approach, the attributes characterizing the user’s behavior and a machine learning algorithm must be selected. The attributes we consider are the locations of body parts in the reference coordinate system (fixed with respect to the environment), the locations of body parts in a body coordinate system (affixed to the user’s body) and the angles between adjacent body parts. Eight machine learning algorithms are compared. The highest classification accuracy of over 95 % is achieved by Support Vector Machine used on the reference attributes and angles.

Falls are common occurrences in elderly people worldwide and may have several adverse consequences, such as physical injuries and psychological distress, leading to decreased functioning and quality of life. Approximately half of the community-living older population experiences fear of falling. The experience of a recent fall is a known cause for the development of fear of falling, but fear of falling is also prevalent in non-fallers: it is plausible that factors related to the process of aging, such as physical frailty, contribute to the development of fear of falling as well. Several studies have indicated that people who are afraid of falling appear to enter a debilitating spiral of loss of confidence, restriction of physical activities and social participation, physical frailty, falls, and loss of independence.

In addition to the adverse consequences of fear of falling for those suffering from it, there are consequences for the public expenditure, because healthcare utilization increases. It is therefore important to reduce fear of falling by reversing the downward spiral by intervening in factors in the spiral, such as increasing physical functioning, or in predictors of those factors, such as improved medication use.

The need for effective falls-prevention strategies is thus evident.

The CONFIDENCE project, funded by the FP7-EU, aims to address these needs by developing an RFID based system composed by a set of tags the end user has to wear, a set of sensors mounted on the walls, a base station and a PDA-like portable device. Communicating with the tags, the sensors on the wall detect the 3D position of the old person. The position is then transmitted to the base station which interpreter and analyse the data allowing to detects abnormalities (falls, dangerous situations, worsening conditions). The portable device is used to configure the system and to raise alarms: the system can automatically perform a phone call to a user defined set of phone numbers.

In the current version, CONFIDENCE is designed to work indoors but the technological results are expected to be very useful for a future outdoor use, extending its protection to virtually any moment of everyday’s life.

CONFIDENCE is mainly addressed to elderly people with no particular problems with their ADL but, the exploitation of the results of the project is considering also alternative scenarios such as helping controlling events into nursing homes or similar structures.

The goal of the system, and thus its name, is to maintain or build back the confidence of mature people fighting the fear of falling. At the same time, this is awaited to lower the costs of hospitalisations lowering public expenditure.

CONFIDENCE project objective consists in the development of innovative domotics solutions for the quality of life of the older persons to promote their independent living, in particular for the identification of anomalous events as falls, loss of conscience or unusual behaviours that could spring from health problems in the elderly ones. Researchers recognised the importance of involving end users in all the phases of the technical development, to guarantee a user-centred and demand-pull driven approach. the user needs and requirement analysis implied an intensive activity of research involving different stakeholders: elderly people, family members, formal and informal caregivers, friends, neighbourhoods, volunteers, policy and decision makers. The compliance of the System with the end-user indications will be constantly monitored through devoted tasks. The psychological, ethical, gender and legal issues associated with the project activities were guaranteed and constantly monitored too. This monitoring is one of the most relevant aspects in current research; in fact, a working prototype will be the result of this multidisciplinary activity.

This paper presents a multi-agent system for the care of elderly people living at home on their own, with the aim to prolong their independence. The system is composed of seven groups of agents providing a reliable, robust and flexible monitoring by sensing the user in the environment, reconstructing the position and posture to create the physical awareness of the user in the environment, reacting to critical situations, calling for help in the case of an emergency, and issuing warnings if unusual behavior is detected. The system has been tested during several on-line demonstrations.

This paper addresses a module within a care system based on daily human behavior extracted from localization data. The proposed method is based on transforming the sequence of posture and spatial information using novel matrix presentation to extract spatial-activity features. Then, outlier detection method is used for classification of individual’s usual and unusual daily patterns regardless of the cause of the problem, be it physical or mental. Initial experiments show that the proposed algorithm successfully discriminates between daily behavior patterns of healthy person and those with health problems.

This paper presents real-time adaptation of alarm detection models in a remote health monitoring system based on user feedback. Real-time adaptation enables systems to fine-tune to the needs and preferences of the user and changes in the environment. This way, the system performance is improved in terms of technical accuracy and subjective user wishes. Two types of alarm detection models were used: (1) models in the form of rules created by domain expert and (2) models induced by machine learning. The problem of adaptation for the rule-based models is defined as Markov decision process. Machine-learning models are adapted by rebuilding the model every time new data is obtained. We tested the adaptation capabilities of the two types of alarm detection models based on their accuracy and time-to-alarm (needed length of possibly critical activity, such as lying on the ground, which causes the models to raise an alarm). Both types of models achieved 90% alarm detection accuracy. The rule-based models decreased time-to-alarm when user-triggered alarms were raised and increased it when the user indicated false alarm. We did not observe this process for the machine-learning models.

We used machine learning to recognize activities from four tags attached to the body. The coordinates of the tags were obtained with Ubisense real-time location system. We determined that the most suitable machine learning algorithm is Random Forest and that the attribute vector is best assembled from a string of attributes belonging to ten consecutive snapshots of tag coordinates. The main contribution of the paper are four methods for improvement of activity recognition, which take into account the time sequence of recognized activities. These methods improve the classification accuracy by 1.66 percentage points.

This paper presents a procedure for developing rule engine by combining machine learning with expert knowledge. This procedure was applied in the domain of classifying human posture based on information of location of body parts. The procedure overcomes in certain measure the problem of over-fitting to non-representative training dataset. Tests show improvement in accuracy of the developed rule engine in comparison to machine learning techniques. Possibilities for improvement and automation of the procedure for rule engine development are discussed at the end of the paper. The presented procedure can be applied to any problem domain in which representative training dataset is not available or is difficult to obtain.

Activity recognition is a fundamental task for analyzing human behaviour and it is usually achieved with a classifier. Although the classifier might be very accurate, it may still produce some false detections and consequently, spurious state transitions – the transitions that do not occur in reality. This paper examines two approaches for reducing spurious activity transitions. The first approach is based on cost-augment grammar classification namely Sequential Grammar-based Classifier, while the second approach uses hidden Markov models. The paper outlines a basic theoretical background of the methods and describes the implementation procedure. The results showed that both methods successfully reduced spurious transitions and improved classification accuracy.

This paper presents a rule engine for classifying human posture according to information about the location of body parts. The rule engine was developed by enriching decision trees with expert knowledge. Results show 5 percentage points improvement in accuracy compared to support vector machines and a significant 11 percentage points compared to decision trees. The incorporation of expert knowledge overcomes the problem of classifier over-fitting observed with classifiers induced with machine learning. Better robustness of the posture classification rule engine is expected in real-life tests in comparison to classifiers induced with machine learning.

CONFIDENCE FP7 project is developing a care system for the elderly based on measurements of locations of radio tags attached to a human body. Posture classification is the basis of reasoning in CONFIDENCE. We first applied a data mining approach to recordings of human behavior in order to develop classifier of posture. However, evaluations with separate training and test set scenario show over fitting. Therefore, we considered enriching classifier by human modification. Posture classifier was developed by human modification of induced decision trees. Besides the 5% improvement in accuracy compared to support vector machines and a significant 11% compared to decision trees, we also expect better robustness of this classifier in real-life tests.

We present the current work on posture and movement recognition in the European FP7 project CONFIDENCE – Ubiquitous Care System to Support Independent Living. CONFIDENCE aims at providing care for the elderly using radio tags attached to the body. The part of the project described in this paper deals with the reconstruction and interpretation of the user’s behavior in order to raise an alarm or issue a warning if a fall or some other unusual activity is detected. We compared several machine learning algorithms and various attributes characterizing the user’s behavior in order to obtain accurate classification of the behavior into predefined activities. The first results show sufficient accuracy so that the final system is expected to substantially help the elderly and enable them to gain confidence in the late years.

Due to the rapid aging of the European population, an effort needs to be made to ensure that the elderly can live longer independently with minimal support of the working-age population. The Confidence project aims to do this by unobtrusively monitoring their activity to recognize falls and other health problems. This is achieved by equipping the user with radio tags, from which the locations of body parts are determined, thus enabling posture and movement reconstruction. In the paper we first give a general overview of the research on fall detection and activity recognition. We proceed to describe the machine learning approach to activity recognition to be used in the Confidence project. In this approach, the attributes characterizing the user’s behavior and a machine learning algorithm must be selected. The attributes we consider are the locations of body parts in the reference coordinate system (fixed with respect to the environment), the locations of body parts in a body coordinate system (affixed to the user’s body) and the angles between adjacent body parts. Eight machine learning algorithms are compared. The highest classification accuracy of over 95 % is achieved by Support Vector Machine used on the reference attributes and angles.

We used machine learning to reconstruct body posture from the coordinates of 12 tags attached to joints. Three sets of attributes were used: (1) tag coordinates and derived attributed in a reference coordinate system, (2) coordinates and derived attributes in a body coordinate system, and (3) angles between body parts. Eight machine learning algorithms were tried. We found that the best attributes are in the reference coordinate system and the best machine learning algorithm is Support Vector Machine.

The Confidence project is designed to support independent living of the elderly. The main goal of the project is reconstructing the user’s posture to detect falls in real time and to detect other abnormalities in behavior. This paper presents a survey of methods for fall detection and activity recognition and discusses how suitable they are for the Confidence project. The presented methods use accelerometers, velocity profiles and visual markers.