Heidi Similä

Automatic feature selection for context recognition in mobile devices

In mobile devices there exist several in-built sensor units and sources which provide data for context reasoning. More context sources can be attached via wireless network connections. Usually, the mobile devices and the context sources are battery powered and their computational and space resources are limited. This sets special requirements for the context recognition algorithms. In this paper, several classification and automatic feature selection algorithms are compared in the context recognition domain. The main goal of this study is to investigate how much advantage can be achieved by using sophisticated and complex classification methods compared with a simple method that can easily be implemented in mobile devices. The main result is that even a simple linear classification algorithm can achieve a reasonably good accuracy if the features calculated from raw data are selected in a suitable way. Usually context recognition algorithms are fitted to a particular problem instance in an off-line manner and modifying methods for on-line learning is difficult or impossible. An on-line version of the Minimum-distance classifier is presented in this paper and it is justified that it leads to considerably higher classification accuracies compared with the static off-line version of the algorithm. Moreover, we report superior performance for the Minimum-distance classifier compared to other classifiers from the view point of computational load and power consumption of a smart phone.

Human Balance Estimation using a Wireless 3D Acceleration Sensor Network

Balance and gait are a consequence of complex coordination between muscles, nerves, and central nervous system structures. The impairment of these functions can pose serious threats to independent living, especially in the elderly. This study was carried out to evaluate the performance of a wireless acceleration sensor network and its capability in balance estimation. The test has been carried out in eight patients and seven healthy controls. The Patients group had larger values in lateral amplitudes of the sensor displacement and smaller values in vertical displacement amplitudes of the sensor. The step time variations for the Patients were larger than those for the controls. A fuzzy logic and clustering classifiers were implemented, which gave promising results suggesting that a person with balance deficits can be recognized with this system. We conclude that a wireless system is easier to use than a wired one and more unobtrusive to the user

Gait analysis and estimation of changes in fall risk factors

Falls are a major problem for older adults. A continuous gait monitoring that provides fall risk assessment would allow timely interventions aiming for preventing falls. The objective of this work was to find out whether gait variables calculated from the acceleration signal measured during walk task in the baseline assessment can predict changes in commonly used fall risk assessment scales after 12 months follow-up. Forty two subjects were measured during walk test with a triaxial acceleration sensor worn on a waist belt at the lower back near the centre of mass. The fall risk was assessed using a test protocol, which included several assessment methods. Gait analysis was able to predict a decline in ABC, BBS and GDS total scores and slower time in STS-5 after twelve-months follow-up. A subsequent study is needed to confirm the models suitability for data recorded in everyday lives.

Collecting a citizen's digital footprint for health data mining

This paper describes a case study for collecting digital footprint data for the purpose of health data mining. The case study involved 20 subjects residing in Finland who were instructed to collect data from registries which they evaluated to be useful for understanding their health or health behaviour, current or past. 11 subjects were active, sending 100 data requests to 49 distinct organizations in total. Our results indicate that there are still practical challenges in collecting actionable digital footprint data. Our subjects received a total of 75 replies (reply rate of 75.0%) and 61 datasets (reception rate of 61%). Out of the received data, 44 datasets (72.1%) were delivered in paper format, 4 (6.6%) in portable document format and 13 (21.3%) in structured digital form. The time duration between the sending of the information requests and reception of a reply was 26.4 days on the average.

Feasibility of digital footprint data for health analytics and services: an explorative pilot study.

BackgroundAs a result of digitalization, data is available about almost every aspect of our lives. Personal data collected by individuals themselves or stored by organizations interacting with people is known as a digital footprint. The purpose of this study was to identify prerequisites for collecting and using digital data that could be valuable for health data analytics and new health services.MethodsResearchers and their contacts involved in a nationwide research project focusing on digital health in Finland were asked to participate in a pilot study on collecting their own personal data from various organizations of their own choice, such as retail chains, banks, insurance companies, and healthcare providers. After the pilot, a qualitative inquiry was adopted to collect semi-structured interview data from twelve active participants in the pilot. Interviews comprised themes such as the experiences of collecting personal data, as well as the usefulness of the data in general and for the participants themselves. Interview data was then analyzed thematically.ResultsEven if the participants had an academic background and were highly motivated to collect and use their data, they faced many challenges, such as quite long delays in the provision of the data, and the unresponsiveness of some organizations. Regarding the usefulness of the acquired personal data, our results show that participants had high expectations, but they were disappointed with the small amount of data and its irrelevant content. For the most part, the data was not in a format that would be useful for health data analytics and new health services. Participants also found that there were actual mistakes in their health data reports.ConclusionsThe study revealed that collecting and using digital footprint data, even by knowledgeable individuals, is not an easy task. As the usefulness of the acquired personal health data mainly depended on its form and usability for services or solutions relevant to an individual, rather than on the data being valuable as such, more emphasis should be placed on providing the data in a reusable form.

Disease state fingerprint for fall risk assessment.

Fall prevention is an important and complex multifactorial challenge, since one third of people over 65 years old fall at least once every year. A novel application of Disease State Fingerprint (DSF) algorithm is presented for holistic visualization of fall risk factors and identifying persons with falls history or decreased level of physical functioning based on fall risk assessment data. The algorithm is tested with data from 42 older adults, that went through a comprehensive fall risk assessment. Within the study population the Activities-specific Balance Confidence (ABC) scale score, Berg Balance Scale (BBS) score and the number of drugs in use were the three most relevant variables, that differed between the fallers and non-fallers. This study showed that the DSF visualization is beneficial in inspection of an individuals significant fall risk factors, since people have problems in different areas and one single assessment scale is not enough to expose all the people at risk.

Video Communication in Remote Rehabilitation and Occupational Therapy Groups

ABSTRACT This paper presents an exploratory study describing the adoption of video-based services for in-home rehabilitation and occupational therapy groups targeted on older adults. The paper focuses on the subjectively experienced value of the new service from the perspectives of health care professionals and older users. The service was evaluated in a 10-month field trial. The qualitative data analysis findings suggest that video-based services can be used successfully in establishing occupational therapy groups, and that the therapeutic goals can also be achieved through video-based group sessions. However, some limitations should be considered in the design of video-mediated group sessions, so, based on this study, a set of recommendations are presented for establishing video-based group work.

Focus Group Evaluation of Scenarios for Fall Risk Assessment and Fall Prevention in Two Countries

Information and communication technologies (ICT) provide means for developing new tools for preventing falls. To enhance adherence to fall prevention interventions, end users need to be engaged from the early phases of the development process. This paper reports the focus group evaluation of five scenarios related to fall risk assessment and fall prevention. There were four focus groups with older adults in both Finland and Spain; 58 participants in all. The most interesting features for the interviewees were usage of intelligent gym equipment, the possibility of peer support and multi-factorial fall risk assessment. The scenario with intelligent gym equipment rose above the others among Finnish participants, while the scenarios were ranked more evenly by Spanish correspondents. The analysis showed that a personal history of falls and a connection to current habits and routines affected the reception of the proposed solutions.

A computationally light classification method for mobile wellness platforms

The core of activity recognition in mobile wellness devices is a classification engine which maps observations from sensors to estimated classes. There exists a vast number of different classification algorithms that can be used for this purpose in the machine learning literature. Unfortunately, the computational and space requirements of these methods are often too high for the current mobile devices. In this paper we study a simple linear classifier and find, automatically with SFS and SFFS feature selection methods, a suitable set of features to be used with the classification method. The results show that the simple classifier performs comparable to more complex nonlinear k-Nearest Neighbor Classifier. This depicts great potential in implementing the classifier in small mobile wellness devices.

Development and user feedback of home technology for preventing falls in older adults

Falls in older adults are an increasing problem in western societies as the population ages. The objective of this paper is to provide an overview of the development process and user feedback of fall prevention exercise software designed for home-dwelling older adults. The developed system was given to 14 older adults for testing at home and testing lasted for up to 16 months. In addition, ten older adults and four health care professionals participated in a feasibility evaluation. Feedback about usability and usefulness are described in the paper. In general, the received feedback was more positive than negative both from the older and professional end users. The requirements for the system to be personalized and tailored according to user profile and needs, and real-time performance monitoring are emphasized in the results.