Human vs. machine : identification of bat species from their echolocation calls by humans and by artificial neural networks

Automated remote ultrasound detectors allow large amounts of data on bat presence and activity to be collected. Processing of such data involves identifying bat species from their echolocation calls. Automated species identification has the potential to provide more consistent, predictable, and potentially higher levels of accuracy than identification by humans. In contrast, identification by humans permits flexibility and intelligence in identification, as well as the incorporation of features and patterns that may be difficult to quantify. We compared humans with artificial neural networks (ANNs) in their ability to classify short recordings of bat echolocation calls of variable signal to noise ratios; these sequences are typical of those obtained from remote automated recording systems that are often used in large-scale ecological studies. We presented 45 recordings (1–4 calls) produced by known species of bats to ANNs and to 26 human participants with 1 month to 23 years of experience in acoustic identification of bats. Humans correctly classified 86% of recordings to genus and 56% to species; ANNs correctly identified 92% and 62%, respectively. There was no significant difference between the performance of ANNs and that of humans, but ANNs performed better than about 75% of humans. There was little relationship between the experience of the human participants and their classification rate. However, humans with <1 year of experience performed worse than others. Currently, identification of bat echolocation calls by humans is suitable for ecological research, after careful consideration of biases. However, improvements to ANNs and the data that they are trained on may in future increase their performance to beyond those demonstrated by humans.

Identification of New Zealand bats in flight from analysis of echolocation calls by artificial neural networks.

Time-expanded and heterodyned echolocation calls of the New Zealand long-tailed Chalinolobus tuberculatus and lesser short-tailed bat Mystacina tuberculata were recorded and digitally analysed. Temporal and spectral parameters were measured from time-expanded calls and power spectra generated for both time-expanded and heterodyned calls. Artificial neural networks were trained to classify the calls of both species using temporal and spectral parameters and power spectra as input data. Networks were then tested using data not previously seen. Calls could be unambiguously identified using parameters and power spectra from time-expanded calls. A neural network, trained and tested using power spectra of calls from both species recorded using a heterodyne detector set to 40 kHz (the frequency with the most energy of the fundamental of C. tuberculatus call), could identify 99% and 84% of calls of C. tuberculatus and M. tuberculata, respectively. A second network, trained and tested using power spectra of calls from both species recorded using a heterodyne detector set to 27 kHz (the frequency with the most energy of the fundamental of M. tuberculata call), could identify 34% and 100% of calls of C. tuberculatus and M. tuberculata, respectively. This study represents the first use of neural networks for the identification of bats from their echolocation calls. It is also the first study to use power spectra of time-expanded and heterodyned calls for identification of chiropteran species. The ability of neural networks to identify bats from their echolocation calls is discussed, as is the ecology of both species in relation to the design of their echolocation calls.

Acoustic identification of 12 species of echolocating bat by discriminant function analysis and artificial neural networks

We recorded echolocation calls from 14 sympatric species of bat in Britain. Once digitised, one temporal and four spectral features were measured from each call. The frequency-time course of each call was approximated by fitting eight mathematical functions, and the goodness of fit, represented by the mean-squared error, was calculated. Measurements were taken using an automated process that extracted a single call from background noise and measured all variables without intervention. Two species of Rhinolophus were easily identified from call duration and spectral measurements. For the remaining 12 species, discriminant function analysis and multilayer back-propagation perceptrons were used to classify calls to species level. Analyses were carried out with and without the inclusion of curve-fitting data to evaluate its usefulness in distinguishing among species. Discriminant function analysis achieved an overall correct classification rate of 79% with curve-fitting data included, while an artificial neural network achieved 87%. The removal of curve-fitting data improved the performance of the discriminant function analysis by 2 %, while the performance of a perceptron decreased by 2 %. However, an increase in correct identification rates when curve-fitting information was included was not found for all species. The use of a hierarchical classification system, whereby calls were first classified to genus level and then to species level, had little effect on correct classification rates by discriminant function analysis but did improve rates achieved by perceptrons. This is the first published study to use artificial neural networks to classify the echolocation calls of bats to species level. Our findings are discussed in terms of recent advances in recording and analysis technologies, and are related to factors causing convergence and divergence of echolocation call design in bats.

Primary and secondary neural networks of auditory prepulse inhibition: a functional magnetic resonance imaging study of sensorimotor gating of the human acoustic startle response

Feedforward inhibition deficits have been consistently demonstrated in a range of neuropsychiatric conditions using prepulse inhibition (PPI) of the acoustic startle eye-blink reflex when assessing sensorimotor gating. While PPI can be recorded in acutely decerebrated rats, behavioural, pharmacological and psychophysiological studies suggest the involvement of a complex neural network extending from brainstem nuclei to higher order cortical areas. The current functional magnetic resonance imaging study investigated the neural network underlying PPI and its association with electromyographically (EMG) recorded PPI of the acoustic startle eye-blink reflex in 16 healthy volunteers. A sparse imaging design was employed to model signal changes in blood oxygenation level-dependent (BOLD) responses to acoustic startle probes that were preceded by a prepulse at 120 ms or 480 ms stimulus onset asynchrony or without prepulse. Sensorimotor gating was EMG confirmed for the 120-ms prepulse condition, while startle responses in the 480-ms prepulse condition did not differ from startle alone. Multiple regression analysis of BOLD contrasts identified activation in pons, thalamus, caudate nuclei, left angular gyrus and bilaterally in anterior cingulate, associated with EMGrecorded sensorimotor gating. Planned contrasts confirmed increased pons activation for startle alone vs 120-ms prepulse condition, while increased anterior superior frontal gyrus activation was confirmed for the reverse contrast. Our findings are consistent with a primary pontine circuitry of sensorimotor gating that interconnects with inferior parietal, superior temporal, frontal and prefrontal cortices via thalamus and striatum. PPI processes in the prefrontal, frontal and superior temporal cortex were functionally distinct from sensorimotor gating.

Acoustic identification of twelve species of echolocating bat by discriminant function analysis and artificial neural networks

We recorded echolocation calls from 14 sympatric species of bat in Britain. Once digitised, one temporal and four spectral features were measured from each call. The frequency-time course of each call was approximated by fitting eight mathematical functions, and the goodness of fit, represented by the mean-squared error, was calculated. Measurements were taken using an automated process that extracted a single call from background noise and measured all variables without intervention. Two species of Rhinolophus were easily identified from call duration and spectral measurements. For the remaining 12 species, discriminant function analysis and multilayer back-propagation perceptrons were used to classify calls to species level. Analyses were carried out with and without the inclusion of curve-fitting data to evaluate its usefulness in distinguishing among species. Discriminant function analysis achieved an overall correct classification rate of 79% with curve-fitting data included, while an artificial neural network achieved 87%. The removal of curve-fitting data improved the performance of the discriminant function analysis by 2 %, while the performance of a perceptron decreased by 2 %. However, an increase in correct identification rates when curve-fitting information was included was not found for all species. The use of a hierarchical classification system, whereby calls were first classified to genus level and then to species level, had little effect on correct classification rates by discriminant function analysis but did improve rates achieved by perceptrons. This is the first published study to use artificial neural networks to classify the echolocation calls of bats to species level. Our findings are discussed in terms of recent advances in recording and analysis technologies, and are related to factors causing convergence and divergence of echolocation call design in bats.

Fine-grained plant classification using convolutional neural networks for feature extraction

We present an overview of the QUT plant classification system submitted to LifeCLEF 2014. This system uses generic features extracted from a convolutional neural network previously used to perform general object classification. We examine the effectiveness of these features to perform plant classification when used in combination with an extremely randomised forest. Using this system, with minimal tuning, we obtained relatively good results with a score of 0:249 on the test set of LifeCLEF 2014.

Improving deep convultional neural networks with unsupervised feature learning

The latest generation of Deep Convolutional Neural Networks (DCNN) have dramatically advanced challenging computer vision tasks, especially in object detection and object classification, achieving state-of-the-art performance in several computer vision tasks including text recognition, sign recognition, face recognition and scene understanding. The depth of these supervised networks has enabled learning deeper and hierarchical representation of features. In parallel, unsupervised deep learning such as Convolutional Deep Belief Network (CDBN) has also achieved state-of-the-art in many computer vision tasks. However, there is very limited research on jointly exploiting the strength of these two approaches. In this paper, we investigate the learning capability of both methods. We compare the output of individual layers and show that many learnt filters and outputs of the corresponding level layer are almost similar for both approaches. Stacking the DCNN on top of unsupervised layers or replacing layers in the DCNN with the corresponding learnt layers in the CDBN can improve the recognition/classification accuracy and training computational expense. We demonstrate the validity of the proposal on ImageNet dataset.

Intelligent prognostics of machinery health utilising suspended condition monitoring data

The ability to forecast machinery failure is vital to reducing maintenance costs, operation downtime and safety hazards. Recent advances in condition monitoring technologies have given rise to a number of prognostic models for forecasting machinery health based on condition data. Although these models have aided the advancement of the discipline, they have made only a limited contribution to developing an effective machinery health prognostic system. The literature review indicates that there is not yet a prognostic model that directly models and fully utilises suspended condition histories (which are very common in practice since organisations rarely allow their assets to run to failure); that effectively integrates population characteristics into prognostics for longer-range prediction in a probabilistic sense; which deduces the non-linear relationship between measured condition data and actual asset health; and which involves minimal assumptions and requirements. This work presents a novel approach to addressing the above-mentioned challenges. The proposed model consists of a feed-forward neural network, the training targets of which are asset survival probabilities estimated using a variation of the Kaplan-Meier estimator and a degradation-based failure probability density estimator. The adapted Kaplan-Meier estimator is able to model the actual survival status of individual failed units and estimate the survival probability of individual suspended units. The degradation-based failure probability density estimator, on the other hand, extracts population characteristics and computes conditional reliability from available condition histories instead of from reliability data. The estimated survival probability and the relevant condition histories are respectively presented as “training target” and “training input” to the neural network. The trained network is capable of estimating the future survival curve of a unit when a series of condition indices are inputted. Although the concept proposed may be applied to the prognosis of various machine components, rolling element bearings were chosen as the research object because rolling element bearing failure is one of the foremost causes of machinery breakdowns. Computer simulated and industry case study data were used to compare the prognostic performance of the proposed model and four control models, namely: two feed-forward neural networks with the same training function and structure as the proposed model, but neglected suspended histories; a time series prediction recurrent neural network; and a traditional Weibull distribution model. The results support the assertion that the proposed model performs better than the other four models and that it produces adaptive prediction outputs with useful representation of survival probabilities. This work presents a compelling concept for non-parametric data-driven prognosis, and for utilising available asset condition information more fully and accurately. It demonstrates that machinery health can indeed be forecasted. The proposed prognostic technique, together with ongoing advances in sensors and data-fusion techniques, and increasingly comprehensive databases of asset condition data, holds the promise for increased asset availability, maintenance cost effectiveness, operational safety and – ultimately – organisation competitiveness.

Identification of factors predicting clickthrough in Web searching using neural network analysis

In this research, we aim to identify factors that significantly affect the clickthrough of Web searchers. Our underlying goal is determine more efficient methods to optimize the clickthrough rate. We devise a clickthrough metric for measuring customer satisfaction of search engine results using the number of links visited, number of queries a user submits, and rank of clicked links. We use a neural network to detect the significant influence of searching characteristics on future user clickthrough. Our results show that high occurrences of query reformulation, lengthy searching duration, longer query length, and the higher ranking of prior clicked links correlate positively with future clickthrough. We provide recommendations for leveraging these findings for improving the performance of search engine retrieval and result ranking, along with implications for search engine marketing

Neural network learning:theoretical foundations

This important work describes recent theoretical advances in the study of artificial neural networks. It explores probabilistic models of supervised learning problems, and addresses the key statistical and computational questions. Chapters survey research on pattern classification with binary-output networks, including a discussion of the relevance of the Vapnik Chervonenkis dimension, and of estimates of the dimension for several neural network models. In addition, Anthony and Bartlett develop a model of classification by real-output networks, and demonstrate the usefulness of classification with a "large margin." The authors explain the role of scale-sensitive versions of the Vapnik Chervonenkis dimension in large margin classification, and in real prediction. Key chapters also discuss the computational complexity of neural network learning, describing a variety of hardness results, and outlining two efficient, constructive learning algorithms. The book is self-contained and accessible to researchers and graduate students in computer science, engineering, and mathematics

A neural-network based flight controller for UASs

This paper presents a nonlinear gust-attenuation controller based on constrained neural-network (NN) theory. The controller aims to achieve sufficient stability and handling quality for a fixed-wing unmanned aerial system (UAS) in a gusty environment when control inputs are subjected to constraints. Constraints in inputs emulate situations where aircraft actuators fail requiring the aircraft to be operated with fail-safe capability. The proposed controller enables gust-attenuation property and stabilizes the aircraft dynamics in a gusty environment. The proposed flight controller is obtained by solving the Hamilton-Jacobi-Isaacs (HJI) equations based on an policy iteration (PI) approach. Performance of the controller is evaluated using a high-fidelity six degree-of-freedom Shadow UAS model. Simulations show that our controller demonstrates great performance improvement in a gusty environment, especially in angle-of-attack (AOA), pitch and pitch rate. Comparative studies are conducted with the proportional-integral-derivative (PID) controllers, justifying the efficiency of our controller and verifying its suitability for integration into the design of flight control systems for forced landing of UASs.

Pulse-coupled neural network performance for real-time identification of vegetation during forced landing

Safety concerns in the operation of autonomous aerial systems require safe-landing protocols be followed during situations where the a mission should be aborted due to mechanical or other failure. On-board cameras provide information that can be used in the determination of potential landing sites, which are continually updated and ranked to prevent injury and minimize damage. Pulse Coupled Neural Networks have been used for the detection of features in images that assist in the classification of vegetation and can be used to minimize damage to the aerial vehicle. However, a significant drawback in the use of PCNNs is that they are computationally expensive and have been more suited to off-line applications on conventional computing architectures. As heterogeneous computing architectures are becoming more common, an OpenCL implementation of a PCNN feature generator is presented and its performance is compared across OpenCL kernels designed for CPU, GPU and FPGA platforms. This comparison examines the compute times required for network convergence under a variety of images obtained during unmanned aerial vehicle trials to determine the plausibility for real-time feature detection.

Estimation of biodiesel properties from chemical composition – An artificial neural network (ANN) approach

Biodiesel, produced from renewable feedstock represents a more sustainable source of energy and will therefore play a significant role in providing the energy requirements for transportation in the near future. Chemically, all biodiesels are fatty acid methyl esters (FAME), produced from raw vegetable oil and animal fat. However, clear differences in chemical structure are apparent from one feedstock to the next in terms of chain length, degree of unsaturation, number of double bonds and double bond configuration-which all determine the fuel properties of biodiesel. In this study, prediction models were developed to estimate kinematic viscosity of biodiesel using an Artificial Neural Network (ANN) modelling technique. While developing the model, 27 parameters based on chemical composition commonly found in biodiesel were used as the input variables and kinematic viscosity of biodiesel was used as output variable. Necessary data to develop and simulate the network were collected from more than 120 published peer reviewed papers. The Neural Networks Toolbox of MatLab R2012a software was used to train, validate and simulate the ANN model on a personal computer. The network architecture and learning algorithm were optimised following a trial and error method to obtain the best prediction of the kinematic viscosity. The predictive performance of the model was determined by calculating the coefficient of determination (R2), root mean squared (RMS) and maximum average error percentage (MAEP) between predicted and experimental results. This study found high predictive accuracy of the ANN in predicting fuel properties of biodiesel and has demonstrated the ability of the ANN model to find a meaningful relationship between biodiesel chemical composition and fuel properties. Therefore the model developed in this study can be a useful tool to accurately predict biodiesel fuel properties instead of undertaking costly and time consuming experimental tests.