Level Of Detection Performance Research Articles

Automatic soil ph level detection using extreme learning machine via image processing

The pH values in the soil, that is, the acid or basic structure of the soil, affects the amounts of nutrients that the plant receives from the soil. For the plant to take the main nutrients in the soil and grow is only possible at suitable pH values. In this paper a novel soil pH level detection method based on optical imaging is proposed. As the level detection algorithm an Extreme Learning Machine (ELM) is used. In the constructed model while the RGB values of the true color soil images and pH index are used as the inputs of ELM the pH level of soil images are used as the output of ELM. In the experimental studies fifty soil sample images obtained from the literature are used. And a significantly high pH level detection performance of 97.5 % is obtained. This result reveals that the proposed method is a significantly important method to determine the pH levels of soil samples and could be a strong alternative to the traditional methods.

A Computationally Efficient Optimal Wigner Distribution in LCT Domains for Detecting Noisy LFM Signals

Recently, Wigner distribution (WD) associated with linear canonical transforms (LCTs) is quickly becoming a promising technique for detecting linear frequency-modulated (LFM) signals corrupted with noises by establishing output signal-to-noise ratio (SNR) inequality model or optimization model. Particularly, the closed-form instantaneous cross-correlation function type of WD (CICFWD), a unified linear canonical Wigner distribution, has shown to be competitive in detecting noisy LFM signals under an extremely low SNR. However, the CICFWD has up to nine LCT free parameters so that it requires a heavy computational load. To improve the efficiency of real-time processing, this paper focuses on the instantaneous cross-correlation function type of WD (ICFWD), which has only six LCT free parameters but is not a special case of the CICFWD. The main advantage of ICFWD is that it could be expected to reduce the computational complexity while maintaining detection performance. This paper first proposes an optimization model to the ICFWD’s output SNR with respect to deterministic signals embedded in additive zero-mean noises. It then deduces the model’s solution to a single component LFM signal added with white noise, leading to the optimal selection strategy on LCT free parameters. Simulation results demonstrate that the ICFWD improves almost a doubling of computing speed in comparison with the CICFWD while sharing the same level of detection performance. To be specific, the computing time of ICFWD in sampling frequencies 5 Hz, 10 Hz, 15 Hz, and 20 Hz is about 0.048 s, 0.111 s, 0.226 s, and 0.392 s, respectively, while 0.075 s, 0.233 s, 0.478 s, and 0.821 s for the computing time of CICFWD; the ICFWD and CICFWD have nearly the same output SNR higher than that of the WD.

Grating Spectrometry and Spatial Heterodyne Fourier Transform Spectrometry: Comparative Noise Analysis for Raman Measurements

The desire for portable Raman spectrometers is continuously driving the development of novel spectrometer architectures where miniaturisation can be achieved without the penalty of a poorer detection performance. Spatial heterodyne spectrometers are emerging as potential candidates for challenging the dominance of traditional grating spectrometers, thanks to their larger etendue and greater potential for miniaturisation. This paper provides a generic analytical model for estimating and comparing the detection performance of Raman spectrometers based on grating spectrometer and spatial heterodyne spectrometer designs by deriving the analytical expressions for the performance estimator (signal-to-noise ratio, SNR) for both types of spectrometers. The analysis shows that, depending on the spectral characteristics of the Raman light and on the values of some instrument-specific parameters, the ratio of the SNR estimates for the two spectrometers () can vary as much as by two orders of magnitude. Limit cases of these equations are presented for a subset of spectral regimes which are of practical importance in real-life applications of Raman spectroscopy. In particular, under the experimental conditions where the background signal is comparable or larger than the target Raman line and shot noise is the dominant noise contribution, the value of is, to a first order of approximation, dependent solely on the relative values of each spectrometer’s etendue and on the number of row pixels in the detector array. For typical values of the key instrument-specific parameters (e.g., etendue, number of pixels, spectral bandwidth), the analysis shows that spatial heterodyne spectrometer-based Raman spectrometers have the potential to compete with compact grating spectrometer designs for delivering in a much smaller footprint (10–30 times) levels of detection performance that are approximately only five to ten times poorer.

Online Localization and Prediction of Actions and Interactions.

This paper proposes a person-centric and online approach to the challenging problem of localization and prediction of actions and interactions in videos. Typically, localization or recognition is performed in an offline manner where all the frames in the video are processed together. This prevents timely localization and prediction of actions and interactions - an important consideration for many tasks including surveillance and human-machine interaction. In our approach, we estimate human poses at each frame and train discriminative appearance models using the superpixels inside the pose bounding boxes. Since the pose estimation per frame is inherently noisy, the conditional probability of pose hypotheses at current time-step (frame) is computed using pose estimations in the current frame and their consistency with poses in the previous frames. Next, both the superpixel and pose-based foreground likelihoods are used to infer the location of actors at each time through a Conditional Random Field enforcing spatio-temporal smoothness in color, optical flow, motion boundaries and edges among superpixels. The issue of visual drift is handled by updating the appearance models, and refining poses using motion smoothness on joint locations, in an online manner. For online prediction of action/interaction confidences, we propose an approach based on Structural SVM that operates on short video segments, and is trained with the objective that confidence of an action or interaction increases as time passes in a positive training clip. Lastly, we quantify the performance of both detection and prediction together, and analyze how the prediction accuracy varies as a time function of observed action/interaction at different levels of detection performance. Our experiments on several datasets suggest that despite using only a few frames to localize actions/interactions at each time instant, we are able to obtain competitive results to state-of-the-art offline methods.

Vocal fatigue induced by prolonged oral reading: Analysis and detection

This article uses prolonged oral reading corpora for various experiments to analyze and detect vocal fatigue. Vocal fatigue particularly concerns voice professionals, including teachers, telemarketing operators, users of automatic speech recognition technology and actors. In analyzing and detecting vocal fatigue, we focused our investigations on three main experiments: a prosodic analysis that can be compared to the results found in related work, a two-class Support Vector Machines (SVM) classifier into Fatigue and Non-Fatigue states using a large set of audio features and a comparison function that estimates the difference in fatigue level between two speech segments using a combination of multiple phoneme-based comparison functions. The experiments on prosodic analysis showed that vocal fatigue was not associated with an increase in fundamental frequency and voice intensity. A two-class SVM classifier using the Paralinguistic Challenge 2010 audio feature set gave an unweighted accuracy of 94.1% for the training set (10-fold cross-validation) and 68.2% for the test set. These results show that the phenomenon of vocal fatigue can be modeled and detected. The comparison function was assessed by detecting increased fatigue levels between two speech segments. The fatigue level detection performance in Equal Error Rate (EER) was 31% using all phonetic segments and yielded EER of 21% after filtering phonetic segments and 19% after filtering phonetic segments and cepstral features. These results show that some phonemes are more sensitive than others to vocal fatigue. These experiments show that the fatigued voice has specific characteristics for prolonged oral reading and suggest the feasibility of vocal fatigue detection.

Decision fusion rules based on multi-bit knowledge of local sensors in wireless sensor networks

For Wireless Sensor Networks (WSNs) with a small quantity of sensors and very low SNR, distributed detection and decision fusion rules based on multi-bit knowledge of local sensors are proposed. At local sensors, observations are quantized to multi-bit local decisions. Three quantification algorithms are investigated, which are based on weight, statistics and redundancy, respectively. Corresponding suboptimal fusion rules at the fusion center are also discussed by approximating the optimal likelihood ratio test. System level detection performance measures, namely probabilities of detection and false alarm, are derived analytically by employing probability theory. Finally, Monte Carlo methods are employed to study the performance of proposed decision fusion rules with parameters such as Rayleigh fading channel and Gaussian noise. Numerical results show that, under non-ideal channel, commonly used schemes based on weight cannot improve the system performance even with a large number and high SNR. Fortunately, schemes based on statistics and redundancy can enhance the system capability when the node is deficient and SNR is low. Furthermore, schemes based on statistics have the best stability among the three schemes, and schemes based on redundancy have the best performance among the three when quantization degree is high.

Performance Analysis of Distributed Detection in a Random Sensor Field

For a wireless sensor network (WSN) with randomly deployed sensors, the performance of the counting rule, where the fusion center employs the total number of detections reported by local sensors for hypothesis testing, is investigated. It is assumed that the signal power decays as a function of the distance from the target. For both the case where the total number of sensors is known and the wireless channels are lossless, and the case where the number of sensors is random and the wireless channels have nonnegligible error rates, the exact system level probability of detection is derived analytically. Some approximation methods are also proposed to attain an accurate estimate of the probability of detection, while at the same time to reduce the computation load significantly. To obtain a better system level detection performance, the local sensor level decision threshold is determined such that it maximizes the system level deflection coefficient.

Distributed detection in a large wireless sensor network

A distributed detection and decision fusion scheme is proposed for a wireless sensor network (WSN) consisting of a large number of sensors. At the fusion center, the total number of detections reported by local sensors are employed for hypothesis testing. Based on the assumption that the received signal power decays as the distance from the target increases, system level detection performance measures, namely probabilities of detection and false alarm, are derived analytically through approximation by using the central limit theorem (CLT). If the number of sensors is sufficiently large, the proposed fusion rule can provide very good system level detection performance, in the absence of the knowledge of local sensors' performances and at low signal to noise ratio (SNR). It is shown that for all the different system parameters we have explored, this fusion rule is equivalent to the optimal fusion rule, which requires much more prior information. To achieve a better system level detection performance, the local sensor level decision threshold should be designed optimally. Numerical methods are employed to find the optimal local sensor level threshold for different sets of system parameters. Guidelines on selecting the optimal local sensor level decision threshold are also provided.

Alcohol intoxication reduces visual sustained attention.

Effects of alcohol intoxication on visual sustained attention were studied using a vigilance task entailing detection of degraded target stimuli. Data were obtained in separate sessions under four ethanol doses, ranging from 0 (placebo) to 1.05 g/kg lean body weight, with periodic maintenance dosing of 0.12 g/kg. Intoxication lowered the overall level of detection performance, and in addition produced dose-related increases in the rate of performance decrement over time. Analysis of performance data using techniques derived from Signal Detection Theory indicated that the decrements were due specifically to alterations in perceptual sensitivity. Examination of eye movements and blinks indicated that the effects of ethanol were not mediated peripherally. Rather, alcohol appears to have deleterious effects on central processing capacity and the availability of capacity over time. The alcohol-related failure of sustained attention may contribute to increased accident risk in tasks requiring continuous performance.

Polarization-switching techniques for coherent optical communications

Several techniques which rely on forcing the polarization state of either the transmitted signal or local oscillator to vary with time in a nonadaptive manner so that polarization-sensitive performance is obtained are discussed. The proposed schemes adopt a novel approach which uses a birefringent medium (for example, high-birefringence single-mode fibers) to implement polarization switching. These techniques require only a single photodetector and give a fixed level of detection performance, with a power penalty of 3 dB relative to heterodyne detection with perfectly matched polarizations. >

Effective masking bandwidths at low frequencies

Determinations of effective masking bandwidths at low frequencies are scarce because of inherent difficulties in such experimentation, including needs for very high signal levels and very quiet listening conditions. The current experimentation was undertaken to estimate signal to noise ratios associated with three levels of detection performance in a two alternative forced choice task at six frequencies between 40 and 1000 Hz. Both signals and noise backgrounds were presented to observers by loudspeaker in an anechoic chamber. Effective masking bandwidths (inferred by the critical ratio method) were observed to decrease from 1000 to 250 Hz, and then to increase substantially at lower frequencies. [Research sponsored by the Applied Technology Laboratory, U.S. Army Aviation Research and Development Command.]

Simple Procedures for Radar Detection Calculations

The literature of radar contains results of Rice, Marcum, Swerling, and Schwartz in several families of curves, which permit radar engineersto estimate the signal energy ratio required for a given level of detectionperformance. The variety of radar problems, however, makes itimpractical to construct curves for all combinations of radar and targetparameters. The concept of detector loss is used here to evaluate lossesattributable to integration and collapsing, with an accuracy of ±0.3 dBon steady targets. This is added to a separate fluctuation loss, modifiedfor diversity effects, to obtain results on all Swerling target modelsand also on partially correlated targets. The accuracy of the combinedlosses is ±0.5 dB for a wide range of detection and false-alarm probabilities.Starting from the basic single-sample detection curves, onlythree additional graphs are needed to find the energy ratio for givendetection performance in any of these cases. Examples are given whichshow the ease with which different radar options may be compared asto performance on an arbitrary type of target.

Detection of Tones in the Absence of External Masking Noise

Sound pressures required for various levels of detection performance, in the absence of external masking, are summarized in iso-detectability contours. Each such contour shows, for frequencies from 125 to 4000 Hz, the SPL of 150-msec tone pulses which yield a specific probability of correct decisions in a two-alternative, temporal, forced-choice psychophysical procedure (2ATFC). Detection in the absence of externally introduced masking noise is found to differ in no qualitative ways from that in the presence of noise. Additional results suggest a spectrum of “internal noise” which is essentially a critical ratio below the sound pressure level of low-detectability (d′ = 1.0) tones in the quiet. Iso-detectability contours for detection performance of 55% and 99% correct in the 2ATFC procedure are found to enclose the range of sound pressure levels currently accepted as audiometric zero (ISO). [Work supported by the National Institutes of Health, Public Health Service, U. S. Department of Health, Education and Welfare.]