False Alarm Research Articles

Reliable detection of generalized convulsive seizures using an off-the-shelf digital watch: A multisite phase 2 study.

The aim of this study was to develop a machine learning algorithm using an off-the-shelf digital watch, the Samsung watch (SM-R800), and evaluate its effectiveness for the detection of generalized convulsive seizures (GCS) in persons with epilepsy. This multisite epilepsy monitoring unit (EMU) phase 2 study included 36 adult patients. Each patient wore a Samsung watch that contained accelerometer, gyroscope, and photoplethysmographic sensors. Sixty-eight time and frequency domain features were extracted from the sensor data and were used to train a random forest algorithm. A testing framework was developed that would better reflect the EMU setting, consisting of (1) leave-one-patient-out cross-validation (LOPO CV) on GCS patients, (2) false alarm rate (FAR) testing on nonseizure patients, and (3) "fixed-and-frozen" prospective testing on a prospective patient cohort. Balanced accuracy, precision, sensitivity, and FAR were used to quantify the performance of the algorithm. Seizure onsets and offsets were determined by using video-electroencephalographic (EEG) monitoring. Feature importance was calculated as the mean decrease in Gini impurity during the LOPO CV testing. LOPO CV results showed balanced accuracy of .93 (95% confidence interval [CI] = .8-.98), precision of .68 (95% CI = .46-.85), sensitivity of .87 (95% CI = .62-.96), and FAR of .21/24 h (interquartile range [IQR] = 0-.90). Testing the algorithm on patients without seizure resulted in an FAR of .28/24 h (IQR = 0-.61). During the "fixed-and-frozen" prospective testing, two patients had three GCS, which were detected by the algorithm, while generating an FAR of .25/24 h (IQR = 0-.89). Feature importance showed that heart rate-based features outperformed accelerometer/gyroscope-based features. Commercially available wearable digital watches that reliably detect GCS, with minimum false alarm rates, may overcome usage adoption and other limitations of custom-built devices. Contingent on the outcomes of a prospective phase 3 study, such devices have the potential to provide non-EEG-based seizure surveillance and forecasting in the clinical setting.

A Review on Securing Wireless Networks against Jamming Attacks

The internet of thins (IoT) framework has helped in design of high speed connected networks which include Wide Area Networks (WANs). Cognitive Radio can be considered to be one of the key enablers of high speed data transfer with evolving generation of wireless networks. Cognitive networks find their applications in military and defence as well with the advent of internet of things and its allied applications in military warfare. Cognitive Radio Networks often share common resources such as bandwidth or spectrum among several users or stations. Due to continued sharing of resources, cognitive networks often come under security attacks, most common of which are jamming and eavesdropping attacks. In this paper, the basics of cognitive IoT networks have been presented with the focus of security aware cognitive IoT networks. Previous work in the allied domain has been discussed along with their salient features. Keywords: Wide Area Network (WAN), Internet of Things (IoT), cognitive radio network (CRN), security aware spectrum assignment, cyber-attacks, false alarm, and throughput.

Lightweight Hardware-Based Cache Side-Channel Attack Detection for Edge Devices (Edge-CaSCADe)

Cache Side Channel Attacks (CSCA) have been haunting most processor architectures for decades now. Existing approaches to mitigation of such attacks have certain drawbacks namely software mishandling, performance overhead, low throughput due to false alarms, etc. Hence, “mitigation only when detected” should be the approach to minimize the effects of such drawbacks. We propose a novel methodology of fine-grained detection of timing-based CSCA using a hardware-based detection module. We discuss the design, implementation, and use of our proposed detection module in processor architectures. Our approach successfully detects attacks that flush secret victim information from cache memory like Flush+Reload, Flush+Flush, Prime+Probe, Evict+Probe, and Prime+Abort, commonly known as cache timing attacks. Detection is on time with minimal performance overhead. The parameterizable number of counters used in our module allows detection of multiple attacks on multiple sensitive locations simultaneously. The fine-grained nature ensures negligible false alarms, severely reducing the need for any unnecessary mitigation. The proposed work is evaluated by synthesizing the entire detection algorithm as an attack detection block, Edge-CaSCADe, in a RISC-V processor as a target example. The detection results are checked under different workload conditions with respect to the number of attackers, the number of victims having RSA,AES and ECC based encryption schemes like ECIES, and on benchmark applications like MiBench and Embench. More than \(98\% \) detection accuracy within \(2\% \) of the beginning of an attack can be achieved with negligible false alarms. The detection module has an area and power overhead of \(0.9\% \) to \(2\% \) and \(1\% \) to \(2.1\% \) for the targeted RISC-V processor core without cache for 1 to 5 counters, respectively. The detection module does not affect the processor critical path and hence has no impact on its maximum operating frequency.

Improvements on Phase Classification Using Atmospheric Melting and Refreezing Energy Based on Soundings

AbstractDetermining precipitation as solid (snow) or liquid (rain) phase is crucial for remote sensing of precipitation. Most phase classification methods rely on near‐surface temperatures. Further attempts to incorporate atmospheric information aloft only achieved mixed level of success, particularly for precipitation with temperature inversion. In our study, we developed a phase classification scheme based on the upper‐level melting energy (ME) and refreezing energy (RE), which is proportional to the area enclosed by the temperature profile and the 0°C isotherm. We performed least squares fitting and linear discriminant analysis to derive phase separation functions using observed surface and sounding data in North America. We provided separation functions for snow conditional probabilities ranging from 30% to 80% for various applications. Compared to a previously published (Probsnow) method, our energy method achieved comparable performance for Type 1 soundings with one near‐surface melting layer, and significantly improves the phase classification scores for Type 2 soundings with an aloft melting layer and a near‐surface refreezing layer. We innovatively combined surface ice‐bulb temperature with the ratio between the ME and RE to represent Type 2 profiles. For Type 2 soundings, our energy method improves the Heidke skill score (HSS) from 0.25 to 0.47 and reduces false alarm rate (FAR) by 0.47 compared to the Probsnow method for 50% threshold. By applying the physically based method, we improved accuracy and HSS, and reduced FAR for more than two thirds of the evaluation stations across the North America. Finally, we tested the application of the new method in satellite snowfall retrievals.

A group target track‐before‐detect approach using two‐stage strategy with maximum‐likelihood probabilistic data association

AbstractIn tracking scenarios involving groups with dense targets, achieving effective data association is challenging due to mutual occlusion and interference among targets. The complexity of the tracking problem is further exacerbated in low‐observable environments by the increase in false alarm rates. The track‐before‐detect (TBD) is an advanced technology for detecting and tracking low‐observable targets, effectively mitigating data association problems by integrating multi‐frame echo data. However, the existing multi‐target TBD algorithms typically assume that the targets are spatially separated and are not suitable for scenarios involving group targets. A group target maximum‐likelihood probabilistic data association (GT‐ML‐PDA) algorithm, based on the concept of TBD, is proposed to track group targets effectively in low‐observable environments. The proposed algorithm divides group target tracking into two stages: group centre trajectory estimation and individual target trajectory estimation. To enhance the performance of the proposed algorithm, two strategies are suggested: modifying the equivalent measurements and extracting independent measurement sets for individual targets. Simulation results demonstrate that the proposed algorithm is capable of effectively tracking numerous individual targets within a group, even in the presence of heavy clutter.

A falls recognition framework based on faster R-CNN and temporal action sequences that can be deployed on home service robots

With the popularity of home service robots (e.g. floor sweepers), robots should be considered to have more features for older adult care. Compared to fixed home monitors with a limited field of view, fall detection with service robots is an ideal solution to keep older adults and disabled people within sight. However, the user’s actions, such as lying on the bed to sleep or slumping on the sofa to rest, cause the traditional fall detection system to generate false alarms, which disrupts the user’s family life. The present work proposed an enhanced faster R-convolutional neural network (CNN) network by incorporating temporal action sequences and fall acceleration computation, demonstrating a reduced misjudgment rate on the service robot platform. Firstly, motion images were captured to obtain the target’s motion area description and action timing at the input stage. Then, the faster R-CNN algorithm was implemented to check the suspected falls further based on the falling acceleration of the detected actions during the training phase. Finally, the proposed temporal action sequences module eliminated the action mistaken for falling. Network training and robotic platform testing demonstrated that the proposed approach distinguished between falls and false alarms, which mitigated the occurrence of false positives. On the service robot platform, experimental results showed that the FAR was 8.19 and processing time was 0.79 s.

Realtime Selection of Optimal Source Parameters Using Ground Motion Envelopes

It is increasingly common for seismic networks to operate multiple independent automatic algorithms to characterise earthquakes in real-time, such as in earthquake early warning (EEW) or even standard network practice. Commonly used methods to select the best solution at a given time are simple and use ad hoc rules. An absolute measure of how well a solution (event origin and magnitude) matches the observations by the goodness-of-fit between the observed and predicted envelopes is a robust and independent metric to select optimal solutions. We propose such a measure that is calculated as a combination of amplitude and cross-correlation fit. This metric can be used to determine when a preferred solution reaches an appropriate confidence level for alerting, or indeed to compare two (or more) different event characterisations directly. We demonstrate that our approach can also be used to suppress false alarms commonly seen at seismic networks. Tests using the 10 largest earthquakes in Switzerland between 2013 and 2020, and events that caused false alarms demonstrate that our approach can effectively prefer solutions with small errors in location and magnitude, and can clearly identify and discard false origins or incorrect magnitudes, at all time scales, starting with the first event characterisation.

Characterizing the performances of different observational precipitation products and their uncertainties over Africa

Validation of observed gridded precipitation datasets sourced from satellites or reanalysis over Africa remains a challenge due to the dearth of in-situ products that can act as a true estimate. To address this gap, this study compares the performance of different precipitation products (gauge, reanalysis, and satellite-based) sourced from the Frequent Rainfall Observations on GridS (FROGS) database over Africa. Satellite products are classified as corrected (incorporating gauge observations into their algorithms) or uncorrected, which implies that temporal variations depend entirely on the satellite. The main aim is to identify regions where precipitation products depict minimal uncertainties, supporting the use of the datasets in understanding precipitation variability in the specific regions. This is achieved by applying the triple collocation approach, which takes advantage of three collocated datasets of the same variable to derive the mean square error without requiring knowledge of the true value. The results show that light precipitation (1–5 mm d−1) was prevalent in most regions of Africa during the study duration (2001–2016). Estimating the spatial distribution of daily precipitation greater than the 90th percentiles suggests that extreme precipitation is mainly detected over the Central Africa region and coastal regions of West Africa, where the majority of uncorrected satellite products show consistent performance. The satellite product CMORPH_V1_RAW shows higher estimates of 90th percentile precipitation among the uncorrected satellite products. The ability of precipitation products to detect rainy or non-rainy days shows that corrected satellite products depict notable agreement for probability of detection and false alarm ratio over most regions of Africa. Overall, better performance is demonstrated by the IMERG products, ARCv2, CHIRPSv2 and PERSIANN_CDRv1r1 (corrected), and GPCC, CPC_v1.0 and REGEN_ALL (gauge) during the study period. Among the reanalysis products, ERA5 datasets shows good performance in estimating daily precipitation over Africa. The optimal maps that show the classification of products in regions where they depict reliable performance can be recommended for various usage by different stakeholders.

High-Speed Spatial–Temporal Saliency Model: A Novel Detection Method for Infrared Small Moving Targets Based on a Vectorized Guided Filter

Infrared (IR) imaging-based detection systems are of vital significance in the domains of early warning and security, necessitating a high level of precision and efficiency in infrared small moving target detection. IR targets often appear dim and small relative to the background and are easily buried by noise and difficult to detect. A novel high-speed spatial–temporal saliency model (HS-STSM) based on a guided filter (GF) is proposed, which innovatively introduces GF into IR target detection to extract the local anisotropy saliency in the spatial domain, and substantially suppresses the background region as well as the bright clutter false alarms present in the background. Moreover, the proposed model extracts the motion saliency of the target in the temporal domain through vectorization of IR image sequences. Additionally, the proposed model significantly improves the detection efficiency through a vectorized filtering process and effectively suppresses edge components in the background by integrating a prior weight. Experiments conducted on five real infrared image sequences demonstrate the superior performance of the model compared to existing algorithms in terms of the detection rate, noise suppression, real-time processing, and robustness to the background.

Pt/Zn-TCPP Nanozyme-Based Flexible Immunoassay for Dual-Mode Pressure-Temperature Monitoring of Low-Abundance Proteins.

Pressure and temperature, as common physical parameters, are important for monitoring human health. In contrast, single-mode monitoring is prone to causing experimental errors. Herein, we innovatively designed a dual-mode flexible sensing platform based on a platinum/zinc-meso-tetrakis(4-carboxyphenyl)porphyrin (Pt/Zn-TCPP) nanozyme for the quantitative monitoring of carcinoembryonic antigen (CEA) in biological fluids with pressure and temperature readouts. The Pt/Zn-TCPP nanozyme with catalytic and photothermal efficiencies was synthesized by means of integrating photosensitizers into porous materials. The flexible sensing system after the antigen-antibody reaction recognized the pressure using a flexible skin-like pressure sensor with a digital multimeter readout, whereas the temperature was acquired via the photoheat conversion system of the Pt/Zn-TCPP nanozyme under 808 nm near-infrared (NIR) irradiation using a portable NIR imaging camera on a smartphone. Meanwhile, the dual-mode flexible sensing system was carried out on a homemade three-dimensional (3D)-printed device. Results revealed that the developed dual-mode immunosensing platform could exhibit good pressure and temperature responses within the dynamic range of 0.5-100 ng mL-1 CEA with the detection limits of 0.24 and 0.13 ng mL-1, respectively. In addition, the pressure and temperature were sensed simultaneously without crosstalk interference. Importantly, the dual-mode flexible immunosensing system can effectively avoid false alarms during the measurement, thus providing great potential for simple and low-cost development for point-of-care testing.

Factors influencing subseasonal predictability of northern Eurasian cold spells

AbstractCold‐air outbreaks have significant impacts on human health, energy consumption, agriculture, and overall well‐being. This study aims to evaluate the effectiveness of Subseasonal‐to‐Seasonal (S2S) models in predicting cold conditions over northern Eurasia, defined here as the lower tercile of weekly mean 2‐metre temperature anomalies. To assess the predictability of these events we use ensemble hindcasts from five prediction systems from the S2S database. Our analysis focuses on identifying the conditions under which the models confidently predict cold temperatures with a high (>0.5) probability 3–4 weeks ahead, which potentially can represent windows of forecast opportunity. We compare the group of forecasts that correctly predicted the events to the group that forecasted events that did not occur in practice (false alarms). Most of the confident forecasts of cold spells, both correct and false alarms, have cold anomalies already in the initial conditions, often in conjunction with either a negative phase of the North Atlantic Oscillation, or Scandinavian Blocking. We find that S2S models tend to overpredict cold temperatures, with false alarms occurring more likely when the forecasts are initialized during a weak polar vortex. Furthermore, most of the confident false alarms receive the signal from the stratosphere rather than following internal tropospheric dynamics. False alarms initialized during the weak polar vortex conditions are more common when the vortex is in a recovery stage and, subsequently, the downward ‐propagating signal is short‐lived in the troposphere. The analysis of forecasts during different Madden–Julian Oscillation (MJO) phases shows that nearly half of all confident correct cold‐temperature forecasts are initialized during an active MJO in phases 6–8. On the other hand, most false alarms occur during phase 3, which we suggest is due to the presence of the Scandinavian Blocking regime in the initial conditions for this phase.

Research on fire detection method of complex space based on multi-sensor data fusion

Efficient and reliable fire detection methods are essential for addressing complex spatial characteristics of older residential communities, which often face poor fire safety conditions, various fire types, and challenges in rescue operations. In response to the problem of lacking comprehensive analysis of multidimensional data in traditional fire detection, which often leads to false alarms and missed alarms, this paper proposes a fire detection method based on multi-sensor data fusion. Various sensors are employed, including smoke sensors, temperature sensors, CO sensors, and flame sensors. These sensors collect multidimensional data, which are fused using improved adaptive filters and fuzzy Bayesian logical reasoning algorithms to establish a fire detection model based on multi-sensor data fusion. To additionally, false alarms and missed detections caused by sensor damage are further solved by exploring hidden spaces. Experimental results demonstrate that the proposed fire detection system based on multi-sensor data fusion effectively analyses fire occurrence and progression in older residential communities, and performs stability, accuracy and real-time property.

Automotive Parts Defect Detection Based on YOLOv7

Various complex defects can occur on the surfaces of small automobile parts during manufacturing. Compared with other datasets, the auto parts defect dataset used in this paper has low detection accuracy due to various defects with large size differences, and traditional target detection algorithms have been proven to be ineffective, which often leads to missing detection or wrong identification. To address these issues, this paper introduces a defect detection algorithm based on YOLOv7. To enhance the detection of small objects and streamline the model, we incorporate the ECA attention mechanism into the network structure’s backbone. Considering the small sizes of defect targets on automotive parts and the complexity of their backgrounds, we redesign the neck portion of the model. This redesign includes the integration of the BiFPN feature fusion module to enhance feature fusion, with the aim of minimizing missed detections and false alarms. Additionally, we employ the Alpha-IoU loss function in the prediction phase to enhance the model’s accuracy, which is crucial for reducing false detection. The IoU loss function also boosts the model’s efficiency at converging. The evaluation of this model utilized the Northeastern University steel dataset and a proprietary dataset and demonstrated that the average accuracy (mAP) of the MBEA-YOLOv7 detection network was 76.2% and 94.1%, respectively. These figures represent improvements of 5.7% and 4.7% over the original YOLOv7 network. Moreover, the detection speed for individual images ranges between 1–2 ms. This enhancement in detection accuracy for small targets does not compromise detection speed, fulfilling the requirements for real-time, dynamic inspection of defects.

Phantom perception as a Bayesian Inference problem - a pilot study.

Tinnitus is the perception of a sound in the absence of an external source. Although the role of the auditory system is well investigates, there is a gap in how multisensory signals are integrated to produce a single percept in tinnitus. Here, we train participants to learn a new sensory environment by associating a cue with a target signal that varies in perceptual threshold. In the test phase, we present only the cue to see if the person perceives an illusion of the target signal. We perform two separate experiments to observe the behavioural and electrophysiological responses to the learning and test phases in (i) healthy young adults and (ii) people with continuous subjective tinnitus and matched controls. We observed that in both parts of the study, the percentage of false alarms was negatively correlated with the 75% detection threshold. Additionally, the perception of an illusion goes together with increased evoked response potential in frontal regions of the brain. In patients with tinnitus, we observe no significant difference in behavioural or evoked response in the auditory paradigm, whereas patients with tinnitus were more likely to report false alarms going together with increased evoked activity during the learning and test phases in the visual paradigm. This impresses the importance of integrity of sensory pathways in multisensory integration and how this process may be disrupted in people with tinnitus. Furthermore, the current study also presents preliminary data supporting evidence that tinnitus patients maybe building stronger perceptual models.

Enhancing disruption prediction through Bayesian neural network in KSTAR

Abstract In this research, we develop a data-driven disruption predictor based on Bayesian deep probabilistic learning, capable of predicting disruptions and modeling uncertainty in KSTAR. Unlike conventional neural networks within a frequentist approach, Bayesian neural networks can quantify the uncertainty associated with their predictions, thereby enhancing the precision of disruption prediction by mitigating false alarm rates through uncertainty thresholding. Leveraging 0D plasma parameters from EFIT and diagnostic data, a Temporal Convolutional Network adept at handling multi-time scale data was utilized. The proposed framework demonstrates proficiency in predicting disruptions, substantiating its effectiveness through successful applications to KSTAR experimental data.

Improving performance of defect detection by setting skewed tolerance and joint tolerances in crimp force monitor

AbstractA Crimp Force Monitor (CFM) detects defects by whether the crimp force curve of the force sensor exceeds the set tolerance, with the goal of low defect miss rate and false alarm rate. In this study, typical defects made by hand are crimped with an automatic crimping machine together with good products to obtain continuous data of the crimping force curves and to process the data. For the peak and area deviations of the crimp force curve, the good products show skew‐normal distribution, and the defective samples show outliers and different deviation intensities in the positive and negative directions. Furthermore, by manually intervening in the batch production of multiple groups of defects and comparing the defect miss rate and false alarm rate of CFM with different tolerance settings, the peak skewed tolerance, joint loose tolerance and joint skewed tolerance tree settings can meet a low defect miss rate and low false alarm rate at the same time. The group test results verify the skew‐normal distribution of crimping force curve deviation rate and the characteristics of deviation intensities, providing reference for the tolerance setting of CFM in automated terminal crimping production.

Matched Stochastic Resonance Enhanced Underwater Passive Sonar Detection under Non-Gaussian Impulsive Background Noise.

Remote passive sonar detection with low-frequency band spectral lines has attracted much attention, while complex low-frequency non-Gaussian impulsive noisy environments would strongly affect the detection performance. This is a challenging problem in weak signal detection, especially for the high false alarm rate caused by heavy-tailed impulsive noise. In this paper, a novel matched stochastic resonance (MSR)-based weak signal detection model is established, and two MSR-based detectors named MSR-PED and MSR-PSNR are proposed based on a theoretical analysis of the MSR output response. Comprehensive detection performance analyses in both Gasussian and non-Gaussian impulsive noise conditions are presented, which revealed the superior performance of our proposed detector under non-Gasussian impulsive noise. Numerical analysis and application verification have revealed the superior detection performance with the proposed MSR-PSNR detector compared with energy-based detection methods, which can break through the high false alarm rate problem caused by heavy-tailed impulsive noise. For a typical non-Gasussian impulsive noise assumption with α=1.5, the proposed MSR-PED and MSR-PSNR can achieve approximately 16 dB and 22 dB improvements, respectively, in the detection performance compared to the classical PED method. For stronger, non-Gaussian impulsive noise conditions corresponding to α=1, the improvement in detection performance can be more significant. Our proposed MSR-PSNR methods can overcome the challenging problem of a high false alarm rate caused by heavy-tailed impulsive noise. This work can lay a solid foundation for breaking through the challenges of underwater passive sonar detection under non-Gaussian impulsive background noise, and can provide important guidance for future research work.

Automated Stand-Alone Bot for Intruder Detection

The Automated Standalone Bot for Intruder Detection System represents a cutting-edge approach to bolstering security measures through the deployment of an intelligent surveillance system. By harnessing sophisticated computer vision techniques, this system is capable of vigilantly monitoring its surroundings and swiftly identifying potential security threats with remarkable accuracy. Through continuous real-time analysis of video feeds and the application of advanced image processing algorithms, the bot can promptly detect any unauthorized individuals or suspicious activities within its designated area of coverage. Upon detecting a potential intruder, the system initiates an immediate response protocol, which may include activating alarms, notifying security personnel, or implementing deterrent measures to deter any further intrusion attempts. What sets this system apart is its ability to continually refine its intruder detection capabilities over time, leveraging machine learning algorithms to adapt to evolving threats and minimize false positives. Moreover, the system demonstrates remarkable discernment by distinguishing between human subjects, animals, and other objects, thereby reducing the occurrence of false alarms and enhancing its overall reliability. This comprehensive approach not only fortifies security measures but also instills a greater sense of confidence in the system's ability to safeguard the premises effectively. Key Words: Automated Standalone Bot, Intruder Detection System, Security Measures, Intelligent System, Computer Vision Techniques, Surveillance, Real-time Analysis, Video Feeds, Image Processing Algorithms, Swift Detection, Unauthorized Individuals, Suspicious Activities, Response Protocol, Alarms, Notification, Deterrent Measures, Machine Learning Algorithms, Adaptation, False Positives, Reliability, Human Subjects, Animals, False Alarms, Premises Security

A silver lining to a busted forecast? Building relationships after the storm through humanising messages

AbstractGrounded in the quiet weather communication typology, we conducted two between‐subjects experiments comparing humanising to organisational voice messages in predicting disaster organisation‐public relationships, publics' message passing intentions, and publics' community resilience perceptions in the U.S. tornado context. Study 1 examines these relationships in the missed event context, where a tornado was not forecasted, but occurred. Study 2 examines these relationships in the false alarm context, where a tornado was forecasted, but did not occur. Results show differing processes across the two studies, with Study 1 results showing direct message effects, but no indirect effects. Study 2 results show indirect effects of the experimental condition on the outcomes via perceptions of conversational human voice. The discussion extends the quiet weather communication typology by theorising how context influences message strategy effectiveness.

Early fault detection for rolling bearings: A meta‐learning approach

AbstractEarly fault detection (EFD) of rolling bearings aims at detecting the early symptoms of faults by monitoring small deviations of health states. Accurate EFD enables predictive maintenance and contributes to the stability of mechanical systems. In recent years, machine learning based methods have shown impressive performance on EFD. Most of the current machine learning‐based methods assume the availability for a large amount of data. However, in practice, the authors may only have a very limited amount of training data, which makes it hard to learn a reliable machine learning model. To address this concern, in this work, the authors propose to tackle EFD via meta learning. Specifically, the authors first formulate EFD as a few‐shot learning problem and then propose to tackle this problem with a metric‐based meta learning method. Furthermore, ensemble learning is further leveraged to improve the detection robustness. For the proposed method, the distribution difference from the working conditions and the bearings are considered. The experimental results on two bearing datasets show that the proposed method can achieve better EFD performance, that is, detecting incipient faults earlier while bringing in lower false alarms, compared with several frequently used EFD methods.