Bernhard Kohn

Embedded vehicle speed estimation system using an asynchronous temporal contrast vision sensor

This article presents an embedded multilane traffic data acquisition system based on an asynchronous temporal contrast vision sensor, and algorithms for vehicle speed estimation developed to make efficient use of the asynchronous high-precision timing information delivered by this sensor. The vision sensor features high temporal resolution with a latency of less than 100 μ s, wide dynamic range of 120 dB of illumination, and zero-redundancy, asynchronous data output. For data collection, processing and interfacing, a low-cost digital signal processor is used. The speed of the detected vehicles is calculated from the vision sensors asynchronous temporal contrast event data. We present three different algorithms for velocity estimation and evaluate their accuracy by means of calibrated reference measurements. The error of the speed estimation of all algorithms is near zero mean and has a standard deviation better than 3% for both traffic flow directions. The results and the accuracy limitations as well as the combined use of the algorithms in the system are discussed.

Vehicle Counting with an Embedded Traffic Data System using an Optical Transient Sensor

In this paper a sensor system for traffic data acquisition is presented. The embedded system, comprising a motion-sensitive optical sensor and a low-cost, low-power DSP, is capable of detecting, counting and measuring the velocity of passing vehicles. The detection is based on monitoring of the optical sensor output within configurable regions of interest in the sensors field-of-view. In particular in this work we focus on the evaluation of the applied vehicle counting algorithm. The verification of the acquired data is based on manually annotated traffic data of 360 minutes length, containing a total of about 7000 vehicles. The counting error is determined for short (3 minutes) and long (60 minutes) time intervals. The calculated error of 99,2% of the short time intervals and 100% of the long time intervals analyzed, remain within commonly recognized margins of 10% and 3% of detection error respectively.

Night-time vehicle classification with an embedded, vision system

The paper presents night-time vehicle classification using an embedded vision system based on an optical transient sensor. This neuromorphic sensor features an array of 128×128 pixels that respond to relative light intensity changes with low latency and high dynamic range. The proposed algorithm exploits the temporal resolution and sparse representation of the data, delivered by the sensor in the data-driven Address-Event Representation (AER) format, to efficiently implement a robust classification of vehicles into two classes, car-like and truck-like, during night-time operation. The classification is based on the extraction of the positions and distances of the vehicles head lights to estimate vehicle width. We present the algorithm, test data and an evaluation of the classification accuracy by comparison of the test data with ground truth from video annotation and reference results from a state-of-the-art ultrasonic/radar-combination reference detector. The results show that the difference in total truck counts with respect to a reference detector and to manually annotated video during nighttime operation under dry and wet road conditions is typically below 6%.

3D detection of periodic limb movements in sleep

The standard polysomnographic method for detecting periodic limb movements in sleep (PLMS) includes measuring the electromyography (EMG) signals from electrodes at the left and right tibialis anterior muscles. This procedure has disadvantages as the cabling affects the patients quality of sleep and the electrodes tend to come off during the night, deteriorating data quality. We used contactless monitoring of body movements by a 3D time-of-flight camera mounted above the bed. Changes in the 3D silhouette indicate motion. Contactless detection of PLMS has several substantial advantages over the EMG and provides more complete and more specific diagnostic data: (1) Motor events caused by other leg muscles than tibialis anterior muscles are fully captured by the 3D method, but missed by EMG. (2) 3D does not react to tonic muscle contractions, where such contractions cause strong deflections in EMG which are annotated as limb movements by most PSG apparatus. Another aspect turned out to be of high practical relevance: Deflections in EMG traces are frequently caused by poor electrode contacts, potentially causing false movement annotations. This can lead to substantial overestimation of the automatically computed PLM index. Contactless sensing completely avoids such problems.The standard polysomnographic method for detecting periodic limb movements in sleep (PLMS) includes measuring the electromyography (EMG) signals from electrodes at the left and right tibialis anterior muscles. This procedure has disadvantages as the cabling affects the patients quality of sleep and the electrodes tend to come off during the night, deteriorating data quality. We used contactless monitoring of body movements by a 3D time-of-flight camera mounted above the bed. Changes in the 3D silhouette indicate motion. Contactless detection of PLMS has several substantial advantages over the EMG and provides more complete and more specific diagnostic data: (1) Motor events caused by other leg muscles than tibialis anterior muscles are fully captured by the 3D method, but missed by EMG. (2) 3D does not react to tonic muscle contractions, where such contractions cause strong deflections in EMG which are annotated as limb movements by most PSG apparatus. Another aspect turned out to be of high practical relevance: Deflections in EMG traces are frequently caused by poor electrode contacts, potentially causing false movement annotations. This can lead to substantial overestimation of the automatically computed PLM index. Contactless sensing completely avoids such problems.

Real-time body motion analysis for dance pattern recognition

This paper presents an algorithm for real-time body motion analysis for dance pattern recognition by use of a dynamic stereo vision sensor. Dynamic stereo vision sensors asynchronously generate events upon scene dynamics, so that motion activities are on-chip segmented by the sensor. Using this sensor body motion analysis and tracking can be efficiently performed. For dance pattern recognition we use a machine learning method based on the Hidden Markov Model. Emphasis is laid on the analysis of the suitability for use in embedded systems. For testing the algorithm we use a dance choreography consisting of eight different activities and a training set of 430 recorded activities performed by 15 different persons. A cross validation on the data reached an average recognition rate of 94%.

Serious Gaming: Enhancing the Quality of Life among the Elderly through Play with the Multimedia Platform SilverGame

The objective of the SilverGame project is to develop a multimedia platform tailored specifically to the needs of the elderly by connecting play-based applications with web-based information and communication services to encourage mental and physical well-being as well as social interaction among the elderly. The project aims to offer users the possibility to enjoy within their homes, leisure activities such as singing, exercise and driving either alone or jointly with other users. They will also be able to access or exchange information using integrated online channels. A prototype multimedia platform and three serious games which can be played using the TV set are being developed by the project partners. The application is operated using an intuitive touch screen interface on a handy tablet computer designed specifically for the elderly target group. The multimedia platform is intended to be offered as a complete system including easy-to-install and easy-to-use hardware. The applications and the services will be available to buy on a modular basis permitting further upgrading at any time.

Sensor Fusion on an Embedded System for Traffic Data Analysis - ETRADA-V System

This paper describes an embedded system for traffic flow analysis based on fused traffic information coming from two different sensors. This can be used as test bed for traffic monitoring. Data is acquired by two sensor systems, Embedded Traffic Data Sensor (TDS) and a CCTV camera. Video data acquired by the CCTV camera is transmitted to an embedded hardware platform, where image processing software runs in order to detect and track objects with the final goal of event detection and event prediction. The system is tested in a real scenario. Experimental results are presented considering the image processing analysis, i.e. object detection and object tracking, and statistical data inferred from previous analysis.

Real-time gesture recognition using bio inspired 3D vision sensor

This paper presents a sensor system for spatiotemporal motion analysis in a 4D representation and real-time pattern analysis based on dynamic stereo vision. The system comprises a dynamic stereo vision sensor, which is sensitive to temporal contrast and therefore reacts to motion such that pixels affected by the scene changes autonomously generate events. The data are completely asynchronous and therefore have ultra-high temporal resolution (1μs@1000 lux) and wide dynamic range (over 120dB). Using this sensor body motion analysis and tracking can be efficiently performed because of the continuous stream of data, which accurately capture changes even in difficult illumination conditions. For dance pattern recognition we use a machine learning method based on the Hidden Markov Model, for a realtime recognition of activities. The whole system has been evaluated on a dance choreography consisting of eight different activities and a training set of 430 recorded activities performed by 15 different persons. Preliminary results show that the proposed system reaches an average recognition rate of 94%.

Ein innovatives, optisches Sensorsystem für die Verkehrsdatenerfassung.

ZusammenfassungDer Artikel beschreibt ein eingebettetes, optisches Sensorsystem, das auf einem speziellen optischen CMOS-Sensor mit analoger Signalvorverarbeitung basiert. Die Daten werden asynchron mit dem so genannten Address-event-Protokoll vom Sensor zur nachgeschalteten digitalen Signalverarbeitung übertragen. Dank der effizienten Datenkodierung und Signalvorverarbeitung, die bewegte Objekte aus der Szene extrahiert, erreicht die Zeitauflösung der digitalen Datenverarbeitung auf einem kostengünstigen digitalen Signalprozessor eine Millisekunde. Die Algorithmen zur Verkehrsdatenerfassung berechnen in Echtzeit die Verkehrsstärke, Fahrgeschwindigkeit, Nettozeitlücke und Belegung auf bis zu vier Fahrstreifen simultan. Der Artikel präsentiert Verkehrsdaten, die an einer vierspurigen Autobahn über vier Tage aufgenommen wurden. Die Zeitverläufe der Verkehrsstärke, der mittleren Fahrgeschwindigkeit sowie der mittleren Nettozeitlücke und der mittleren Belegung wurden in Fünfminutenintervallen ausgewertet. Der Fehler in der Geschwindigkeitsschätzung liegt unter 3 %, die Genauigkeit der Fahrzeugzählung liegt, bei Tageslicht und Normalbedingungen, über 97 %.SummaryAn embedded vision system based on a specialized CMOS optical sensor with on-chip analogue signal pre-processing is described. The sensor signal is transmitted to digital processing units via asynchronous address-event protocol. Using the efficient data coding and signal pre-processing concept, moving objects can be extracted from the visual scene with a temporal resolution of 1 millisecond. The traffic data acquisition algorithms compute traffic flow, vehicle velocity, separation and lane occupancy in real-time on up to four lanes simultaneously. We present traffic data that has been acquired continuously over four days at a test site at a four lane highway. The error of the velocity estimation is below 3 %, the precision of the vehicle flow measurement is better than 97 % under normal daylight conditions.

Contactless 3D detection of respiratory effort

Respiratory effort is a major feature for detection and classification of apneas in polysomnography. Presently, somnologists apply flow sensors and/or rip belts at the thorax and abdomen for this purpose, causing practical problems with the montage and re-adjustment during the night and disturbing patients´ sleep. Contactless measurements would be a desirable alternative. We utilized a 3D time-of-flight camera to monitor respiratory-related chest movements to decipher epochs of normal breathing and apnea. Time-synchronized comparisons of 3D measurements of chest movements due to respiration to signals from rip belts and nasal airflow proved that the 3D sensor provided equivalent results. This new technique could support the diagnosis of sleep apnea and Cheyne-Stokes breathing. It simplifies the procedure, saves personnel capacity, improves data quality and releases the burden to the patient by replacing body-mounted sensors and cabling.