UREDT: Unsupervised Learning Based Real-Time Footfall Event Detection Technique in Seismic Signal

Anchal, Sahil; Mukhopadhyay, Bodhibrata; Kar, Subrat

doi:10.1109/lsens.2017.2787611

Cited by 22 publications

(8 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Acoustic Event Detection: Artificial neural networks have been applied to acoustic event classification [8,9,19,41] which includes among others footstep detection. Also footstep detection and person identification using geophones has been studied before [3,21,30], however only in experiments in a controlled environment, not on embedded devices or using additional structural information. Artificial neural networks have been recently applied to seismic event detection [29].…”

Section: Related Workmentioning

confidence: 99%

“…Here, several challenges need to be addressed. Multiple footstep detectors using geophones have been proposed [3,30] but have not been shown to distinguish well between footsteps and seismic events [27] or require further structural information [21]. Convolutional neural networks have shown to be good signal processing tools for classification of acoustic [19] as well as seismic sources [32].…”

Section: Classification With Time Distributed Processingmentioning

confidence: 99%

See 1 more Smart Citation

Event-triggered natural hazard monitoring with convolutional neural networks on the edge

Meyer

Farei-Campagna

Pasztor

et al. 2019

Proceedings of the 18th International Conference on Information Processing in Sensor Networks

View full text Add to dashboard Cite

In natural hazard warning systems fast decision making is vital to avoid catastrophes. Decision making at the edge of a wireless sensor network promises fast response times but is limited by the availability of energy, data transfer speed, processing and memory constraints. In this work we present a realization of a wireless sensor network for hazard monitoring based on an array of eventtriggered single-channel micro-seismic sensors with advanced signal processing and characterization capabilities based on a novel co-detection technique. On the one hand we leverage an ultra-low power, threshold-triggering circuit paired with on-demand digital signal acquisition capable of extracting relevant information exactly and efficiently at times when it matters most and consequentially not wasting precious resources when nothing can be observed. On the other hand we utilize machine-learning-based classification implemented on low-power, off-the-shelf microcontrollers to avoid false positive warnings and to actively identify humans in hazard zones. The sensors' response time and memory requirement is substantially improved by quantizing and pipelining the inference of a convolutional neural network. In this way, convolutional neural networks that would not run unmodified on a memory constrained device can be executed in real-time and at scale on low-power embedded devices. A field study with our system is running on the rockfall scarp of the Matterhorn Hörnligrat at 3500 m a.s.l. since 08/2018.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Classification With Time Distributed Processingmentioning

confidence: 99%

Event-triggered natural hazard monitoring with convolutional neural networks on the edge

Meyer

Farei-Campagna

Pasztor

et al. 2019

Proceedings of the 18th International Conference on Information Processing in Sensor Networks

View full text Add to dashboard Cite

show abstract

“…The image classifier architecture is selected from the large pool of available image classifiers. For micro-seismic data, three different classifiers will be pre-selected: (i) a footstep detector based on manually selected features (standard deviation, kurtosis and frequency band energies) using a linear support vector 5 machine (LSVM) similar to the detector used in (Anchal et al, 2018), (ii) a seismic event classifier adopted from (Perol et al, 2018) and (iii) an acoustic event classifier. We reimplemented the first two algorithms based on the information from the respective papers.…”

Section: Classifier Selectionmentioning

confidence: 99%

“…Due to its simplicity, a popular filtering technique for event detection is to use short-term/long-term average triggering (STA/LTA) (Withers et al, 1998). This is often used in the analysis of unstable slopes (Colombero et al, 2018;20 Levy et al, 2011), is available in commercial data loggers (Geometrics, 2018) and can be used to detect footsteps (Anchal et al, 2018). Due to it's inherent simplicity, STA/LTA cannot reliably discriminate seismic activity from external (unwanted) influence factors such as noise from human beings, wind, rain or hail without manually supervising and intervening with the detection process.…”

mentioning

confidence: 99%

Interactive Figures

Meyer¹

2018

Preprint

View full text Add to dashboard Cite

Natural hazards, e.g. due to slope instabilities, are a significant risk for the population of mountainous regions.Monitoring of micro-seismic signals can be used for process analysis and risk assessment. However, these signals are subject to external influences, e.g anthropogenic or natural noise. Successful analysis depends strongly on the capability to cope with such external influences. For correct slope characterization it is thus important to be able to identify, quantify and take these influences into account. 5In long-term monitoring scenarios manual identification is infeasible due to large data quantities demanding accurate automated analysis methods. In this work we present a systematic strategy to identify multiple external influences, characterize their impact on micro-seismic analysis and develop methods for automated identification. We apply the developed strategy to a real-word, multi-sensor, multi-year micro-seismic monitoring experiment on the Matterhorn Hörnliridge (CH). We present a convolutional neural network for micro-seismic data to detect external influences originating in mountaineers, a major un-10 wanted influence, with an error rate of less than 1 %, which is 3x lower than comparable algorithms. Moreover, we present an ensemble classifier for the same task obtaining an error rate of 0.79 % and an F1 score of 0.9383 by using images and micro-seismic data. Applying the classifiers to the experiment data reveals that approximately 1/4 of events detected with an event detector are not due to seismic activity but due to anthropogenic mountaineering influences and that time periods with mountaineer activity have a 9x higher event rate. Due to these findings we argue that a systematic identification of external 15 influences, like presented in this paper, is a prerequisite for a qualitative analysis.

show abstract

“…Due to its simplicity, a popular filtering technique for event detection is to use short-term/long-term average triggering (STA/LTA) (Withers et al, 1998). This is often used in the analysis of unstable slopes (Colombero et al, 2018;Levy et al, 2011), is available in commercial data loggers (Geometrics, 2018) and can be used to detect footsteps (Anchal et al, 2018). Due to it's inherent simplicity, STA/LTA cannot reliably discriminate seismic activity from external (unwanted) influence factors such as noise from human beings, wind, rain or hail without manually supervising and intervening with the detection process.…”

mentioning

confidence: 99%

Response to RC1

Meyer¹

2018

Preprint

View full text Add to dashboard Cite

Natural hazards, e.g. due to slope instabilities, are a significant risk for the population of mountainous regions. Monitoring of micro-seismic signals can be used for process analysis and risk assessment. However, these signals are subject to external influences, e.g anthropogenic or natural noise. Successful analysis depends strongly on the capability to cope with such external influences. For correct slope characterization it is thus important to be able to identify, quantify and take these influences into account. In long-term monitoring scenarios manual identification is infeasible due to large data quantities demanding accurate automated analysis methods. In this work we present a systematic strategy to identify multiple external influences, characterize their impact on micro-seismic analysis and develop methods for automated identification. We apply the developed strategy to a real-word, multi-sensor, multi-year micro-seismic monitoring experiment on the Matterhorn Hörnliridge (CH). We present a convolutional neural network for micro-seismic data to detect external influences originating in mountaineers, a major unwanted influence, with an error rate of less than 1 %, which is 3x lower than comparable algorithms. Moreover, we present an ensemble classifier for the same task obtaining an error rate of 0.79 % and an F1 score of 0.9383 by using images and micro-seismic data. Applying the classifiers to the experiment data reveals that approximately 1/4 of events detected with an event detector are not due to seismic activity but due to anthropogenic mountaineering influences and that time periods with mountaineer activity have a 9x higher event rate. Due to these findings we argue that a systematic identification of external influences, like presented in this paper, is a prerequisite for a qualitative analysis. Copyright statement. TEXT 1 Introduction Rock-slope failures and landslides pose a significant risk for humans and infrastructure (Glade et al., 2005). Identifying and monitoring potential slope instabilities is therefore of great importance. Micro-seismic analysis can be used to assess the stability and characteristics of slopes in various environments (Spillmann et al., 2007) and as a result is beneficial for two main applications: (i) process analysis and scientific understanding and (ii) natural hazard warning systems. With the advent of 1

show abstract

UREDT: Unsupervised Learning Based Real-Time Footfall Event Detection Technique in Seismic Signal

Cited by 22 publications

References 10 publications

Event-triggered natural hazard monitoring with convolutional neural networks on the edge

Event-triggered natural hazard monitoring with convolutional neural networks on the edge

Interactive Figures

Response to RC1

Contact Info

Product

Resources

About