Unacceptable Rate Of False Alarms Research Articles

Deep autoencoder for false positive reduction in handgun detection

In an object detection system, the main objective during training is to maintain the detection and false positive rates under acceptable levels when the model is run over the test set. However, this typically translates into an unacceptable rate of false alarms when the system is deployed in a real surveillance scenario. To deal with this situation, which often leads to system shutdown, we propose to add a filter step to discard part of the new false positive detections that are typical of the new scenario. This step consists of a deep autoencoder trained with the false alarm detections generated after running the detector over a period of time in the new scenario. Therefore, this step will be in charge of determining whether the detection is a typical false alarm of that scenario or whether it is something anomalous for the autoencoder and, therefore, a true detection. In order to decide whether a detection must be filtered, three different approaches have been tested. The first one uses the autoencoder reconstruction error measured with the mean squared error to make the decision. The other two use the k-NN (k-nearest neighbors) and one-class SVMs (support vector machines) classifiers trained with the autoencoder vector representation. In addition, a synthetic scenario has been generated with Unreal Engine 4 to test the proposed methods in addition to a dataset with real images. The results obtained show a reduction in the number of false positives between 22.5% and 87.2% and an increase in the system’s precision of 1.2%-47% when the autoencoder is applied.

Robust Epileptic Seizure Detection on Wearable Systems with Reduced False-Alarm Rate.

Epilepsy affects more than 50 million people and ranks among the most common neurological diseases worldwide. Despite advances in treatment, one-third of patients still suffer from refractory epilepsy. Wearable devices for real-time patient monitoring can potentially improve the quality of life for such patients and reduce the mortality rate due to seizure-related accidents and sudden death in epilepsy. However, the majority of employed seizure detection techniques and devices suffer from unacceptable false-alarm rate. In this paper, we propose a robust seizure detection methodology for a wearable platform and validate it on the Physionet.org CHB-MIT Scalp EEG database. It reaches sensitivity of 0.966 and specificity of 0.925, and reducing the false-alarm rate by 34.7%. We also evaluate the battery lifetime of the wearable system including our proposed methodology and demonstrate the feasibility of using it in real time for up to 40.87 hours on a single battery charge.

Detection and Analysis of Near-Surface Explosions on the Kola Peninsula

Seismic and infrasonic observations of signals from a sequence of near-surface explosions at a site on the Kola Peninsula have been analyzed. NORSAR’s automatic network processing of these events shows a significant scatter in the location estimates and, to improve the automatic classification of the events, we have performed full waveform cross-correlation on the data set. Although the signals from the different events share many characteristics, the waveforms do not exhibit a ripple-for-ripple correspondence and cross-correlation does not result in the classic delta-function indicative of repeating signals. Using recordings from the ARCES seismic array (250 km W of the events), we find that a correlation detector on a single channel or three-component station would not be able to detect subsequent events from this source without an unacceptable false alarm rate. However, performing the correlation on each channel of the full ARCES array, and stacking the resulting traces, generates a correlation detection statistic with a suppressed background level which is exceeded by many times its standard deviation on only very few occasions. Performing f-k analysis on the individual correlation coefficient traces, and rejecting detections indicating a non-zero slowness vector, results in a detection list with essentially no false alarms. Applying the algorithm to 8 years of continuous ARCES data identified over 350 events which we confidently assign to this sequence. The large event population provides additional confidence in relative travel-time estimates and this, together with the occurrence of many events between 2002 and 2004 when a temporary network was deployed in the region, reduces the variability in location estimates. The best seismic location estimate, incorporating phase information for many hundreds of events, is consistent with backazimuth measurements for infrasound arrivals at several stations at regional distances. At Lycksele, 800 km SW of the events, as well as at ARCES, infrasound is detected for most of the events in the summer and for few in the winter. At Apatity, some 230 km S of the estimated source location, infrasound is detected for most events. As a first step to providing a Ground Truth database for this useful source of infrasound, we provide the times of explosions for over 50 events spanning 1 year.