Książki

This book covers the results of a study concerning systems for healthcare-oriented monitoring of elderly persons. It is focused on the methods for processing data from impulse-radar sensors and depth sensors, aimed at localisation of monitored persons and estimation of selected quantities informative from the healthcare point of view. It includes mathematical descriptions of the considered methods, as well as the corresponding algorithms and the results of their testing in a real-world context. Moreover, it explains the motivations for developing healthcare-oriented monitoring systems and specifies the real-world needs which may be addressed by such systems.
The healthcare systems, all over the world, are confronted with challenges implied by the ageing of population and the lack of adequate recruitment of healthcare professionals. Those challenges can be met by developing new technologies aimed at improving the quality of life of elderly people and at increasing the efficiency of public health management. Monitoring systems may contribute to this strategy by providing information on the evolving health status of independently-living elderly persons, enabling healthcare personnel to quickly react to dangerous events. Although these facts are generally acknowledged, such systems are not yet being commonly used in healthcare facilities and households. This may be explained by the difficulties related to the development of technological solutions which can be both acceptable for monitored persons and capable of providing healthcare personnel with useful information. The impulse-radar sensors and depth sensors, considered in this book, have a potential for overcoming those difficulties since they are not cumbersome for the monitored persons - if compared to wearable sensors - and do not violate the monitored person's privacy - if compared to video cameras. 
Since for safety reasons the level of power, emitted by the radar sensors, must be ultra-low, the task of detection and processing of signals is a research challenge which requires more sophisticated methods than those developed for other radar applications. This book contains descriptions of new Bayesian methods, applicable for the localisation of persons by means of impulse-radar sensors, and an exhaustive review of previously published ones. Furthermore, the methods for denoising, regularised numerical differentiation and fusion of data from impulse-radar sensors and depth sensors are systematically reviewed in this book. On top of that, the results of experiments aimed at comparing the performance of various data-processing methods, which may serve as guidelines for related future projects, are presented.

Non-invasive Monitoring of Elderly Persons: Systems Based on Impulse-Radar Sensors and Depth Sensors

1.        Introduction to healthcare-oriented monitoring of persons

1.1.        Objectives of healthcare-oriented monitoring

1.2.        Systems for healthcare-oriented monitoring

1.2.1.         Monitoring techniques

1.2.2.         Commercially available monitoring systems

1.3.        Semantic and mathematical modelling of monitoring systems

1.3.1.         Semantic modelling of monitoring systems

1.3.2.         Mathematical modelling of monitoring systems

1.4.        Overview of measurement-data processing in healthcare-oriented monitoring systems

1.4.1.         Localisation of persons by means of impulse-radar sensors

1.4.2.         Localisation of persons by means of depth sensors

1.4.3.         Denoising and differentiation of persons' movement trajectories

1.4.4.         Fusion and postprocessing of data from impulse-radar sensors and depth sensors

2.        Localisation of persons by means of impulse-radar sensors - basic methods

2.1.        Methods for extraction of signal from impulse-radar data

2.1.1.         Mathematical model of impulse-radar data

2.1.2.         Method based on arithmetic averaging

2.1.3.         Method based on exponential averaging

2.1.4.         Method based on singular-value decomposition

2.2.        Methods for estimation of impulse-radar signal parameters

2.2.1.         Mathematical model of impulse-radar data

2.2.2.         Method based on time-domain template matching

2.2.3.         Method based on frequency-domain template matching

2.2.4.         Method based on maximum-envelope matching

2.3.        Methods for estimation of two-dimensional trajectories

2.3.1.         Methods for smoothing of distance trajectories

2.3.2.         Methods for conversion of two distance trajectories into movement trajectory

2.4.        Chapter conclusions

3.        Localisation of persons by means of impulse-radar sensors - advanced methods

3.1.        Principles of Bayesian inference

3.1.1.         Bayesian methods for measurand reconstruction

3.1.2.         Key role of a priori information

3.1.3.         Recent applications of Bayesian inference in data processing

3.2.        Extraction of signal from impulse-radar data

3.2.1.         Method #1

3.2.2.         Method #2

3.2.3.         Method #3

3.3.        Estimation of impulse-radar signal parameters

3.3.1.         Dealing with varying number of echoes

3.3.2.         Dealing with fixed number of echoes

3.4.        Estimation of two-dimensional trajectories

3.4.1.         Smoothing of distance trajectories

3.4.2.         Conversion of two distance trajectories into movement trajectory

4.        Localisation of persons by means of impulse-radar sensors - comparison of methods

4.1.        Extraction of signal from impulse-radar data

4.1.1.         Methodology of numerical experimentation based on semi-synthetic data

4.1.2.         Numerical experiments based on semi-synthetic data

4.1.3.         Numerical experiments based on real-world data

4.1.4.         Section conclusions

4.2.        Estimation of impulse-radar signal parameters

4.2.1.         Methodology of numerical experimentation based on semi-synthetic data

4.2.2.         Numerical experiments based on semi-synthetic data

4.2.3.         Methodology of numerical experimentation based on real-world data

4.2.4.         Numerical experiments based on real-world data

4.2.5.         Section conclusions

4.3.        Estimation of two-dimensional trajectories

4.3.1.         Methodology of numerical experimentation based on real-world data

4.3.2.         Numerical experiments based on real-world data

4.3.3.         Section conclusions

5.        Denoising and differentiation of persons' movement trajectories - basic methods

5.1.        Mathematical formulation of denoising and differentiation problems

5.1.1.         Mathematical model of measurement data

5.1.2.         Outline of studied methods for denoising and differentiation

5.1.3.         Classification of studied methods for denoising and differentiation

5.2.        Scalar methods for denoising and differentiation

5.2.1.         Method based on Savitzky-Golay filtering

5.2.2.         Method based on Kalman filtering

5.2.3.         Finite-difference methods for differentiation

5.3.        Vector polynomial methods for denoising and differentiation

5.3.1.         Method based on Tikhonov regularisation

5.3.2.         Method based on regularisation by constraining total variation

5.3.3.         Method based on singular value decomposition

5.3.4.         Method based on Landweber's algorithm

5.3.5.         Method based on CGLS algorithm

5.4.        Vector non-polynomial methods for denoising and differentiation

5.4.1.         Method based on regularisation by constraining number of basis functions

5.4.2.         Method based on Tikhonov regularisation

5.5.        Catalogue of studied methods for regularised numerical differentiation

6.        Denoising and differentiation of persons' movement trajectories - advanced methods

6.1.        Implementation of parameter-optimisation strategies

6.1.1.         Strategy based on discrepancy principle

6.1.2.         Strategy based on generalised cross-validation

6.1.3.         Strategy based on L-curve

6.1.4.         Strategy based on normalised cumulative periodogram

6.1.5.         Strategy based on Stein's unbiased risk estimator

6.2.        Comparative study of parameter-optimisation strategies

6.2.1.         Methodology of experimentation

6.2.2.         Optimisation of degree of approximating polynomial

6.2.3.         Optimisation of parameters of Kalman filter

6.2.4.         Optimisation of differentiation step

6.2.5.         Optimisation of Tikhonov regularisation parameters

6.2.6.         Optimisation of constraint on total variation

6.2.7.         Optimisation of number of retained components of SVD

6.2.8.         Optimisation of number of iterations

6.2.9.         Optimisation of number of basis functions

6.2.10.     Summary of experimentation results

7.        Fusion of data from impulse-radar sensors and depth sensors

7.1.        Basic methods for fusion of data

7.1.1.         Method based on ordinary least-squares estimator

7.1.2.         Method based on total least-squares estimator

7.1.3.         Method based on Kalman filter

7.2.        Advanced methods for fusion of data

7.2.1.         Method based on maximum-a-posteriori estimator

7.2.2.         Method based on maximum-likelihood estimator

7.2.3.         Method based on particle filter

7.3.        Methodology of numerical experimentation based on real-world data

7.3.1.         Acquisition of measurement data

7.3.2.         Criteria for performance evaluation

7.4.        Numerical experimentation based on real-world data

7.4.1.         Straight-line trajectories and square-shaped trajectories

7.4.2.         Serpentine trajectories

7.5.        Chapter conclusions

8.        Gait analysis

8.1.        Objectives and methods of gait analysis

8.1.1.         Methods for characterisation of human gait

8.1.2.         Applicability of spatiotemporal gait analysis in healthcare practice

8.1.3.         Overview of techniques for estimation of self-selected walking speed

8.1.4.         Overview of techniques for estimation of other spatiotemporal gait parameters

8.1.5.         Overview of techniques for gait analysis based on depth sensors

8.2.        Comparison of methods for estimation of walking speed based on impulse-radar sensors and depth sensors

8.2.1.         Methodology of experimentation

8.2.2.         Results of experimentation

9.        Fall detection

9.1.        Techniques for fall detection

9.1.1.         Overview of techniques for fall detection

9.1.2.         Techniques for fall detection based on depth sensors

9.1.3.         Performance of techniques for fall detection

9.2.        Comparison of methods for fall detection based on depth sensors

9.2.1.         Methodology of experimentation

9.2.2.         Results of experimentation

10.     Ethical issues related to monitoring of persons

10.1.    General scheme of making morally significant decisions

10.2.    Bioethical principles and values

10.3.    Technoscientific recommendations

10.3.1.     Measurements

10.3.2.     Data processing

11.     Conclusion

11.1.    Final remarks

11.2.    Further research

References

Index