Peter Christ

Performance analysis of the nRF24L01 ultra-low-power transceiver in a multi-transmitter and multi-receiver scenario

Low-power Wireless Sensor Networks (WSN) are used in various energy constraint applications in industry, medicine, and human motion monitoring. In this paper we analyze the transmission performance of a WSN in a static indoor test setup consisting of 5 to 14 transmitters and up to four receivers. In particular, we consider the nRF24L01 transceiver that implements the proprietary ANT protocol and operates in the 2.4 GHz ISM band. We conduct tests in over 500 configurations with different message frequencies, packet sizes and number of transmitters. Our tests experimentally point out the tradeoff between rate and package length regarding the number of lost packets. Moreover, we show that using multiple receivers reduces the amount of consecutively lost packages. The obtained results provide valuable insights for drawing conclusions for an application specific device configuration as well as redundancy mechanisms for maximizing the information throughput.

Fine-Grained Prediction of Cognitive Workload in a Modern Working Environment by Utilizing Short-Term Physiological Parameters

In this paper we present a method to predict cognitive workload during the interaction with a tablet computer. To set up a predictor that estimates the reflected self-reported cognitive workload we analyzed the information gain of heart rate, electrodermal activity and user input (touch) based features. From the derived optimal feature set we present a Gaussian Process based learner that enables fine-grained and short term detection of cognitive workload. Average inter-subject accuracy in 10-fold cross validation is 74.1 % for the fine-grained 5-class problem and 96.0 % for the binary class problem.

A respiration sensor for a chest-strap based wireless body sensor

In this paper we present a respiration sensor suitable for an integration into a wireless body sensor worn around the chest. The thorax expansion and contraction during in- and exhalation is captured using a force-sensing resistor. Based on the captured thoracic movements, the breaths are determined with a peak detection algorithm. For evaluation, a treadmill experiment with five subjects was conducted using an ergospirometry system as a reference. Overall, an average deviation of -0.32±0.68 min-1 in the respiration rate between the ergospirometry and our sensor was observed. In general, the captured thoracic movements showed breaths as distinctive oscillations, but in some cases a non-optimal pressure transfer between thorax and sensor was observed. Therefore, a mechanical housing mechanism was developed. A comparison of our construction with a respiratory inductance plethysmography (RIP)-based sensor shows a close relationship with the captured thoracic movements during normal and deep respiration.

Pareto-optimal signal processing on low-power microprocessors

Miniaturised wireless body sensors equipped with low-power microcontrollers are used in various energy-constrained applications. The signal-processing algorithms often require running in real-time on a low computational and memory budget. In this paper we present a framework for the exploration of the design space of resource-efficient signal processing suitable for embedded processors. Using a velocity estimation algorithm for an athlete, we show which configurations of the algorithm perform best in respect to classification accuracy and runtime. Altering the sampling frequency, the feature combination, the classifier (Artificial Neural Network (ANN), Decision Tree (DT)), or the classifiers parametrisation, we obtained 15 Pareto-optimal configurations out of 1008 simulations. The highest classification accuracy of 93.92% was obtained using an ANN, and required 22422 clock cycles per classification. The lowest cycle count of 204 was obtained with a DT configuration which resulted in 84.66 % accuracy.

Robust estimation of physical activity by adaptively fusing multiple parameters

Raising the awareness of being physically active by utilizing wearable body sensors has become a popular research topic. Recent approaches combine physical and physiological information to obtain a precise prediction of a person;s physical activity ratio. However, the error in the determination of physical activity due to invalid physiological values that are resulting from underlying signal disturbances, has so far not been considered. We therefore present a robust measure of activity that fuses accelerometer data, heart rate and other personalized features, and is adaptively responding to missing physiological sensor data. To set up the model, we make use of regression analysis (MARS). Our findings indicate the need for considering signal quality when estimating physical activity. The predictive model shows close agreement (R2 = 0.97) to the reference from indirect calorimetry, even if the physiological information is partly corrupted.

Design-space exploration of the configurable 32 bit VLIW processor CoreVA for signal processing applications

In this paper we present the results of a design-space exploration for a classification algorithm with respect to the inherent parallelism of the CoreVA CPU. The CoreVA is a configurable VLIW processor which has been mainly designed for energy-constrained applications. Energy-efficient signal-processing is essential for real-time applications on wireless body sensors (WBSs). Using a velocity-estimation algorithm for a runner as an example, we show which hardware and algorithm configurations perform best in respect to classification accuracy, runtime and energy consumption. We obtained 9 Pareto-optimal configurations out of 504 simulations. The highest classification accuracy of 93.4% requires 34687 clock cycles and has an energy consumption of 1.559 μJ. The lowest energy requirements of 0.015μJ per classification are observed with a Pareto-optimal configuration at 76.3% accuracy. The three-issue VLIW configuration shows the best results with respect to the area-energy trade-off.

Identification of Athletes During Walking and Jogging Based on Gait and Electrocardiographic Patterns

We propose a biometric method for identifying athletes based on information extracted from the gait style and the electrocardiographic (ECG) waveform. The required signals are recorded within a non-clinical acquisition setup using a wireless body sensor attached to a chest strap with integrated textile electrodes. Our method combines both sources of information to allow identification despite severe intra-subjects variations in the gait patterns (walking and jogging) and motion related artefacts in the ECG patterns. For identification we use features extracted in time and frequency domain and a standard classifier. Within a treadmill experiment with 22 subjects we obtained an accuracy of 98.1 % for velocities from 3 to 9 km/h. On a second data set consisting of 9 subjects and two sessions of recording, our method achieved 93.8 % despite variations in the patterns due to reapplying the body sensor and an increased velocity (up to 11 km/h).

Detailed Estimation of Cognitive Workload with Reference to a Modern Working Environment

In modern industry, employees are confronted with ever more complex working tasks. As a consequence, cognitive workload of the employees rises. This makes automatic estimation of cognitive workload a key subject of research. Such an estimate would enable adaptive Human-Machine Interaction that could be used to fit the employees’ workload accordingly to their needs. In this work, a tablet interaction study is presented that is designed to induce cognitive workload. Supervised machine learning methods are used to estimate the induced cognitive workload based on features taken from heart rate, electrodermal activity and user interaction (touch input). Ground truth data is obtained from the subjects’ self-reported cognitive workload. Inter-subject accuracy of the best learner is 74.1% for the detailed 5-class problem and 96.0% for the simplified binary problem.