Enrico Köppe

Binarized-BLSTM-RNN based Human Activity Recognition

High computational complexity hinders the widespread usage of neural networks, especially in mobile devices, which are often the basis of fine-grained localization technology for ubiquitous health monitoring, context awareness, and indoor location tracking. In this paper, we present a binarized recurrent neural network whose weight parameters, input, and intermediate hidden layer output signals, are all binary-valued, and require only basic bit logic for the evaluation and training process. The proposed Binarized Long Short-Term Memory Network (B-BLSTM-RNN) is especially suitable for resource-constrained environments since it replaces either floating or fixed-point arithmetic with significantly more efficient bitwise operations. The model is based on a bidirectional Long Short-Term Memory Recurrent Neural Network (BLSTM-RNN). Designed to take contextual information into account, the network can process data gathered from different positions, resulting in a system, thats invariant to transformations and distortions of the input patterns. During the forward pass, the B-BLSTM drastically reduce memory size and accesses, and replace most arithmetic operations with bit-wise operations, which is expected to substantially improve power-efficiency. The binarized network is simple, accurate, efficient, and works on challenging gesture recognition tasks using raw MEM data. To validate the effectiveness of the network we conduct three sets of experiments. We achieved a classification accuracy with a the proposed network of about 90% which only 2% less than the full-precision network. We also compare our method with recent methods and outperform these methods by large margins on the conducted datasets.

An advanced method for pedestrian dead reckoning using BLSTM-RNNs

Location estimation and navigation, especially on smartphones has shown great progress in the past decade due to its low cost and ability to work without additional infrastructure. However, a challenge is the positioning, both in terms of step detection, step length approximation as well as heading estimation, which must be accurate and robust, even when the use of the device is varied in terms of placement or orientation. In this paper, we propose a scheme for retrieving relevant information to detect steps and to estimate the correct step length from raw inertial measurement unit (IMU) data. This approach uses Bidirectional Long Short-Term Memory Recurrent Neural Networks (BLSTM-RNNs). Designed to take contextual information into account, the network can process data gathered from different positions, resulting in a system, which is invariant with respect to transformation and distortions of the input patterns. An experimental evaluation on a dataset produced from 10 individuals demonstrates that this new approach achieves significant improvements over previous attempts and increase the current state-of-the-art results even in the presence of variations and degradations. We achieved a mean classification rate of 98.5% and a standard deviation of 0.70 for 10000 different test sequences and an average error of 1.45% regarding the step length. Thus is the best result on the task gathered in the experiments compared with competing techniques.

Radio-based multi-sensor system for person tracking and indoor positioning

Sensor based person tracking is a challenging topic. The main objective is positioning in areas without GPS connection, i.e. indoors. A research project is carried out at BAM, Federal Institute for Materials Research and Testing, to develop and to validate a multi-sensor system for 3D localization. It combines body motion sensing and a guard system for the tracking and recording of the status of persons. The so named BodyGuard system was designed for sensor-based monitoring and radio-based transmission of the movement of a person. Algorithms were developed to transform the sensor data into a spatial coordinate. This paper describes how the BodyGuard system operates, which main components were used in the system, how the individual sensor data are converted into 3D motion data, with which algorithms the individual sensors are processed, how individual errors are compensated and how the sensor data are merged into a 3D Model. Final objective of the BodyGuard system is to determine the exact location of a person in a building, e.g. during fire-fighting operations.

Self-calibration-method for an inertial navigation system with three 3D sensors

Inertial Navigation Systems with three 3D sensors are used to localize moving persons. The accuracy of the localization depends on the quality of the sensor data of the multi-sensor system. In order to improve the accuracy, a self-calibration process based on the automatic 3D calibration was developed. Based on the calibration procedure of the accelerometer (ACC) and the magnetic field sensor (MAG), the additional integration of the gyroscope (GYRO) leads to a reduction of the indoor positioning error. This improves both the approximation for the accelerometer and the magnetic field sensor so that the standard deviation of a single sensor is minimized. A new calibration procedure of the gyroscope and the accuracy improvement of the localization of a moving person are presented.

Automatic 3D calibration for a multi-sensor system

A major current focus in multi-sensor systems for indoor localisation is how to calibrate the sensors. In this study, we introduce a new method for this task. By using the datafusion of the combined sensor data, we were able to eliminate the errors of an individual sensor. The challenging matter was the connection of the different sensors for a continuous calibration. The necessary data for the calibration were given due to the natural movement of a person. The examination of this values offers an automatic re-calibration without using a given calibration sequence. This finding is promising for improving the calculation of the exact position of a person, who is moving free in a room.

Erprobung eines Messsystems mit Datenbus und dezentraler Datenspeicherung für den Einsatz bei FallprüfungenValidation of a Measuring System with Data Bus and Local Data Recording for Application in Drop Tests

Zusammenfassung Zur Untersuchung der Sicherheit von Behältern für den Transport und die Lagerung von Gefahrstoffen und -gütern werden Fallprüfungen durchgeführt. Die Aufpralldynamik und die strukturmechanischen Einwirkungen auf den Behälter werden mittels Beschleunigungsaufnehmern und Dehnungsmessstreifen erfasst. Dabei kommen derzeit Vielkanalmesssysteme zum Einsatz, die eine Verkabelung jeder einzelnen Messstelle und somit das Mitfallen eines Kabelbaums erfordern, wodurch Probleme bei der Versuchsvorbereitung und Durchführung entstehen. Die Verwendung eines Messsystems mit Datenbus und dezentraler Datenspeicherung bietet diesbezüglich einen vielversprechenden Lösungsansatz.

Application of the Inertial Navigation System 3D-Self-Calibration-Method for the Minimization of the Measurement Uncertainty

For the accuracy of inertial navigation systems for indoor localization it is important to get high quality sensor data of the multi-sensor system. This can be realized using high quality sensors or the developed 3D-self-calibration-method for low cost sensors. Based on the calibration procedure of the accelerometer (ACC) and the magnetic field sensor (MAG), the additional integration of the gyroscope (GYRO) leads to a reduction of the indoor positioning error. This improves both the approximation for the accelerometer, the magnetic field sensor and the gyroscope so that the standard deviation of a single sensor is minimized. There are errors in the whole system. To determine these error sources it is important to define the measurement uncertainty. In this paper it is presented that the measurement uncertainty can be reduced by the application of the developed 3D-self-calibration method.

Einsatz geführter Wellen für die Ultraschallprüfung

Zusammenfassung Der Einsatz geführter Wellen für die zerstörungsfreie Prüfung mit Ultraschall eröffnet neue Möglichkeiten, räumlich ausgedehnte Bauteile mit begrenzter Zugänglichkeit auf ihre Integrität zu prüfen und gewinnt daher zunehmend an Bedeutung. Dieser Artikel behandelt die physikalischen Grundlagen der Schallausbreitung. Deren Verständnis bildet die Grundlage für die Entwicklung geeigneter Prüfsysteme. An Beispielen werden verschiedene Möglichkeiten zur Simulation der Schallausbereitung vorgestellt. Aktuelle Lösungsansätze zur Prüfung von plattenförmigen Strukturen und von Rohrleitungen werden beschrieben, wobei besonders auf die Sensortechnik und die speziellen Anforderungen an die Prüfhardware eingegangen wird. Abstract Guided waves are widely used for non-destructive testing using ultrasound. Recently, the method has become increasingly important for integrity tests of spatially extended components with limited accessibility. This article discusses the basic physics of the sound propagation of guided waves. Their understanding forms the basis for the successful development of adapted inspection systems. Examples for simulating the wave propagation using different methods are presented. Current approaches for the inspection of plate-like structures and pipelines are described with focus on sensor technology and the specific requirement on the test hardware.