Paweł Mazurek

Acquisition and preprocessing of data from infrared depth sensors to be applied for patients monitoring

A methodology for acquisition and preprocessing of real-world data, necessary for the development of classification algorithms dedicated to fall detection, is proposed. Raw measurements are acquired by means of infrared depth sensors. Their preprocessing consists of two main operations: extraction of the silhouette from the depth image and estimation of the coordinates of the center of that silhouette and its magnitude. The algorithms of classification under development are supplied with sequences of those parameters, corresponding to the consecutive images acquired by a depth sensor.

Deep learning classifier for fall detection based on IR distance sensor data

The goal of research is the fall detection in elderly residents based on infra red depth sensor measurements. Our attention is focused on statistical properties as generalization. The effectiveness of discriminative statistical classifiers (multilayer perceptron) is improved by addition of feature selection block by Gram-Schmidt orthogonalization, which determines the ranking of the features, and NPCA block, which transforms the raw data into a nonlinear manifold and reduces the dimensionality of the data. Performance of our system measured in terms of sensitivity is 92% and precision is 93%, which means it can be used for real life applications.

Application of k Nearest Neighbors Approach to the fall detection of elderly people using depth-based sensors

This paper presents the methodology of the elderly peoples fall detection using the k Nearest Neighbors (kNN) as the decision making module. The problem of the data acquisition by the depth sensors and feature selection for this task is introduced. The classification problem is discussed. The decision making algorithm and its parameters are briefly described. Experimental results based on data collected in the laboratory are presented and commented. The paper is concluded with future prospects of the approach and its possible modifications.

Application of naïve Bayes classifier in fall detection systems based on infrared depth sensors

A novel solution of the fall detection problem, based on the use of infrared depth sensors, is proposed. A methodology for acquisition of real-world data and their preprocessing is presented. The procedures for feature generation, preprocessing and selection are described. The naïve Bayes classifier is designed for the selected features and its performance is evaluated using a data set consisting of 144 sequences representative of 72 falls and 72 other human activities.

Selected algorithms for measurement data processing in impulse-radar-based system for monitoring of human movements

The importance of research on new technologies that could be employed in care services for elderly and disabled persons is highlighted. Advantages of impulse-radar sensors, when applied for non-intrusive monitoring of such persons in their home environment, are indicated. Selected algorithms for the measurement data preprocessing - viz. the algorithms for clutter suppression and echo parameter estimation, as well as for estimation of the twodimensional position of a monitored person - are proposed. The capability of an impulse-radar- based system to provide some application-specific parameters, viz. the parameters characterising the patients health condition, is also demonstrated.

Neural network classifier for fall detection improved by Gram-Schmidt variable selection

The paper describes results of research on the fall detection in elderly residents based on infra red depth sensor measurements. We present the methodology of data acquisition, preprocessing and the feature extraction. Multilayer perceptron is used for classification. In order to improve the classifier generalization feature selection block by Gram-Schmidt orthogonalization is added. It determines the ranking of the features and enables to reduce the dimensionality of the data. Performance of our system measured in terms of sensitivity is 92% and precision is 93%, which means it can be used for real life applications.

Semiconductor optical amplifiers and Raman amplification for 1310-nm dense wavelength division multiplexed transmission

Abstract. We investigate the utilization of semiconductor optical amplifiers (SOAs) and quantum-dot laser-based Raman amplifiers in high-capacity dense wavelength division multiplexed (DWDM) 1310-nm transmission systems. Performed simulations showed that in a 10×40  Gbit/s system, the utilization of a single Raman amplifier in a back-propagation scheme can extend the maximum error-free (bit error rate <10−9) transmission distance by approximately 25 km in comparison with the same system utilizing only an SOA used as a preamplifier. We successfully applied a Raman amplifier in an 8×2×40  Gbit/s 1310-nm polarization multiplexed (PolMux) DWDM transmission over 25 km. Conducted experiments showed that the utilization of a Raman amplifier in this system leads to 4-dB improvement of the average channel sensitivity in comparison to the same system utilizing SOAs. This sensitivity improvement can be translated into a higher power budget. Moreover, lower input optical power in a system utilizing a Raman amplifier reduces the four-wave mixing interactions. The obtained results prove that Raman amplification can be successfully applied in 1310-nm high-capacity transmission systems, e.g., to extend the reach of 400G and 1T Ethernet systems.

Use of kinematic and mel-cepstrum-related features for fall detection based on data from infrared depth sensors

Abstract A methodology for acquisition and preprocessing of measurement data from infrared depth sensors, when applied for fall detection, combined with several approaches to the classification of those data, is proposed. Data processing is initiated with extraction of the silhouette from the depth image and estimation of the coordinates of the center of that silhouette. Next, two groups of features to be applied for a fall/non-fall classification are extracted: kinematic features (various statistics defined on the position, velocity and acceleration trajectories of the monitored person) and mel-cepstrum-related features (components of the mel-cepstrum obtained by means of an unconventional set of mel-filters). Finally, the utility of these features in fall detection is assessed using three classification algorithms − viz. support vector machine, artificial neural network, and naive Bayes classifier − trained and tested on two datasets consisting of, respectively, 160 data sequences (representative of 80 falls and 80 other human behaviours) and 264 data sequences (representative of 132 falls and 132 other human behaviours). The application of the combination of the kinematic and mel-cepstrum-related features yields highly accurate classification results − all classifiers achieved, depending on the dataset, 98.6–100% and 93.9–97.7% sensitivity. Thus, infrared depth sensors can be promising tools for unobtrusive fall detection. They provide data which can be in various ways preprocessed to form a basis for reliable fall detection. Appropriate selection of the feature sets directly affects the reliability of unobtrusive monitoring systems, and − indirectly − the quality of life of the monitored persons.

Healthcare-Oriented Characterisation of Human Movements by Means of Impulse-Radar Sensors and by Means of Accelerometric Sensors.

This paper is devoted to the healthcare-oriented characterisation of the human movements by means of the accelerometric and impulse-radar sensors – the sensors that may be employed in care services for monitoring of elderly and disabled persons. Characterisation of the movements in terms of the so-called self-selected walking velocity can be used by the medical and healthcare personnel to assess the overall health status of a monitored person. The quality of the characterisation, based on the measurement data from accelerometric and impulse-radar sensors, has been assessed in a series of real-world experiments which involved the estimation of the instantaneous and mean walking velocity of a person moving according to predefined patterns. Some indicators of uncertainty of the velocity estimation, determined with respect to assumed predefined velocity values, have been used for comparison of the performance of both types of sensors. The experiments have shown that impulse-radar sensors enable one to estimate the mean walking velocity more accurately than the accelerometric sensors: the estimates obtained on the basis of data from the latter sensors are affected by larger bias and are more widely spread around their mean values.

Fusion of measurement data from impulse-radar sensors and depth sensors when applied for patients monitoring

The research reported in this paper is related to the fusion of measurement data from the impulse-radar sensors and infrared depth sensors, i.e. significantly different sensors that may be employed for unobtrusive monitoring of elderly and disabled persons. The performance of monitoring systems, based on both types of sensors - when used separately and when used in combination - has been compared in a series of experiments which involved the tracking of a moving person. The results have shown that the combined use of sensors, followed by the adequate fusion of measurement data, alleviates the problems occurring if only single-type sensors are used for monitoring, viz. the bias and dispersion of the estimates decreases and the blind spots in the monitored area disappear. Since the sequence of the position estimates can be used for derivation of many healthcare-related parameters, e.g. mean walking velocity, the application of data fusion may considerably increase the reliability of the unobtrusive monitoring of elderly and disabled persons.