Cosamed

1 Introduction

In 2000, approximately one million people were already supported and monitored by pendant or pull-cord alarms linked to centralised control facilities [7]. A large group of people is not able to live in their familiar surroundings anymore without having regular surveillance and care from trained medical staff or family members. Suffering from dementia, Parkinson or amyotrophia, just to mention a few makes an independent life impossible. A problem for the independent living of elderly people are the risks to fall in their homes, which can result in severe fractures. Additionally they might not be able to get up or call for help on their own. Caregivers as well as family members would like to have up to date information on the condition of their patients or relatives. Under these presumptions, human factors and ethics are an important factor to consider: On the one hand, those people are kept under surveillance during their daily activities intruding their privacy, on the other hand this monitoring can give them the opportunity to live a free and independent life on their own [14].

1.1 Population Prospects

Costs for health care and the number of ill and old people have been increasing steadily over the last decades. Until the year 2050, the worldwide population aged 60 years and over will increase from 11% in 2006 to 22% [3, 2].

For the more developed regions (Europe, Northern America, Australia, New Zealand and Japan) this value is expected to rise from 20% in 2006 to 32% until 2050 (fig 1.1). In Austria in 2006, 23% of the people were over 60 . In 2050, 37% of the total population will be over 60, 35% of those being over 80 years. The support ratio - the ratio of people aged 15 to 64 years to the people aged 65 or over - will decrease from to 2 by 2050. While this indicates the quality of the health care systems and an improvement in the state of health of the more developed countries, it also means more age-related diseases and raise the question of long-term care. This development will become a great challenge for the health care systems of the more developed countries [6].

A survey conducted in 1999 in the UK with 11,500 users involved, showed that the expected saving due to the introduction of smart homes for telecare would be around £7,100,000. Translating these results to the UK as a whole the savings would be around 7.7 billion over a 10–year period [5].

2 Related Work

A variety of health smart homes have been presented [15], [13], [14], [10] that use magnetic switches, IR sensors and environmental sensors like thermometers or hygrometers. Switches and motion detectors provide reliable data, however the information is often incomplete and too coarse. These systems have been enhanced by incorporating physiological data including weight, blood oxygen saturation , skin temperature or heart rate. As these are worn in addition to normal clothing, they can limit the mobility and can be hard to operate if manual triggering of measuring is required. It is believed that computer vision can be used to develop a sensing system that overcomes most of the problem of these types of hardware, while allowing to continually monitor the state of health without the problem of constraining people during their activities.

Beginning with assisting people in a rather small and limited domain of their living space, it was shown that people with dementia can profit from having a hand wash monitoring system [11]: A computer vision system mounted over the sink gives information to the patient about already having their hands washed, leaving the water running or not having used soap. Figure 2 illustrates the tracking of hands and objects. The system is designed to be extensible, with the hand-washing agent just as an example that is: (1) easily to define and simple to model, (2) complex enough to pose problems to patients with dementia, (3) safe for clinical trials, without many concerns about privacy.

---

---

Nait-Charif and McKenn [12] have shown that fall detection and activity summarisation can be achieved with a single overhead-mounted camera per room. The people are modelled and tracked as ellipses using a particle filter. The authors claim that representing persons as ellipses yields to a representation that is rich enough to allow detection relevant actions such as standing, sitting or falling, while it is coarse enough to allow different body poses and clothing to be tracked. Figure 3 shows an example with the motion trajectories and the inactivity zones marked.

---

---

Fleck et. al. [8] presented a distributed camera tracking system prototype based on a network of smart cameras to perform geo-referenced tracking and activity recognition. Each camera independently perform tracking and object recognition. The results are transmitted to a centralised server. It handles initialization of camera nodes, and hand over of tracking targets to other cameras, as well as collection and storage of the tracking and activity recognition data to a database system. This approach enables the system to scale well even with a large number of cameras attached. The results have been visualized using a 3D reconstruction of the environment in Google Earth and in a web application. Figure 4 shows the visualization in Google Earth.

---

---

2.1 EU Projects

The EU is funding several projects under ETEN and IST that deal with supporting people with aging in place and helping handicapped people.

The ParkService project (ETEN-517382) [1] will validate the European market for a new service supporting people with Parkinson’s disease in their own homes. ParkService combines the secure exchange of video and messages through a simple television interface with a unique mobility aid for people with Parkinson’s disease(see Figure 2.1).

Ambient Assisted Living for older people under various scenarios is the objective in the PERSONA (Perceptive Spaces Promoting Independent Aging) (IST-2005- 2.6.2) [16]. The scenarios range from giving support in activities of daily living like showering and administration of drugs. To in home safety with notification if white goods are not being turned off after usage and fall detection. Additionally people are supported when going out of their homes by a personal wireless device. Information such as weather forecasts and appointment reminders should help people to live an autonomous life.

---

---

---

---

The goal of the INHOME project is to provide the means for improving the quality of life of elderly people at home, by developing generic technologies for managing their domestic ambient environment, comprised of white goods, entertainment equipment and home automation systems with the aim to increase their autonomy and safety. Among them are Home Environment Simple Management, Tasks scheduling, Flexible AV streams handling, Household appliances flexible access and Activity Monitoring [9].

The problem of delayed calls to medical service is addressed in the EMERGE Project, by supporting elderly people with emergency monitoring and prevention [4]. The approach is to use ambient and unobtrusive sensors to monitor activity, location, and vital data.

2.2 Corporate interest

Companies are already selling products and medical sensors for telecare and telemonitoring applications. However vision related approaches are not on the market yet.

Figure 7 shows the layout of the Vigil Dementia System developed by Vigil Health Solutions. Incidents are automatically reported to the appropriate caregiver via a pager or wireless phones [17].

---

---

The LifeShirt is a sensor embedded garment developed by VivoMetrcis [18], measuring heart rate and posture and activity level using a three-axis accelerometer (Figure 8).

---

---

The full report is available as pdf

References

 

 

Test videos

 

 

For evaluation, we build a set of artificial scenes for fall detection. From 4 camera positions, we take videos from our test person walking by, using the computer and sitting on a chair. At a certain point, the test person is falling down and lying for some seconds. After that, the person is getting up again and walking on. We define the time where the person in the video is falling and lying as the emergency action and the remaining rest as usual behaviour.

8 scenarios are taken in parallel from 4 synchronized cameras. The data set may be requested via mail