Brian A. Bucci

Microphone Array Analysis Methods Using Cross-Correlations

Due to civilian noise complaints and damage claims, there is a need to establish an accurate record of impulse noise generated at military installations. Current noise monitoring systems are susceptible to false positive detection of impulse events due to wind noise. In order to analyze the characteristics of noise events, multiple channel data methods were investigated. A microphone array was used to collect four channel data of military impulse noise and wind noise. These data were then analyzed using cross-correlation functions to characterize the input waveforms. Four different analyses of microphone array data are presented. A new value, the min peak correlation coefficient, is defined as a measure of the likelihood that a given waveform originated from a correlated noise source. Using a sound source localization technique, the angle of incidence of the noise source can be calculated. A method was also developed to combine the four individual microphone channels into one. This method aimed to preserve the correlated part of the overall signal, while minimizing the effects of uncorrelated noise, such as wind. Lastly, a statistical method called the acoustic likelihood test is presented as a method of determining if a signal is correlated or not.Copyright

Nonlinear Control Algorithm for Improving Settling Time in Systems With Friction

A nonlinear control algorithm that greatly reduces settling time in precision instruments with rolling element bearings is proposed. Reductions of 80.5%-87.4% in settling time were achieved when settling to within 3-100 nm of the commanded position. Final settling of such systems is typically impacted by the nonlinearity in the pre-rolling friction regime, which manifests as a hysteretic stiffness. Consequently, the integral term in the controller can take a long time to respond. In this paper, a nonlinear integral action settling algorithm is presented. The nonlinear integral gain takes the form of a Dahl friction model. Since the integral gain mimics hysteretic stiffness, the output of the integral control term is instantaneously set to a large value after each direction change, greatly improving settling response. A nearly first-order error dynamic results, which has a user-definable time constant. Before the algorithm can be implemented, the Coulomb friction and initial contact stiffness in the Dahl model must be experimentally determined for the stage. A sensitivity study is performed on the initial contact stiffness, which was found in other works to dictate the stability of the algorithm.

Performance of artificial neural network-based classifiers to identify military impulse noise

Noise monitoring stations are in place around some military installations to provide records that assist in processing noise complaints and damage claims. However, they are known to produce false positives (by incorrectly attributing naturally occurring noise to military operations) and also fail to detect many impulse events. In this project, classifiers based on artificial neural networks were developed to improve the accuracy of military impulse noise identification. Two time-domain metrics--kurtosis and crest factor--and two custom frequency-domain metrics--spectral slope and weighted square error-were inputs to the artificial neural networks. The classification algorithm was able to achieve up to 100% accuracy on the training data and the validation data, while improving detection threshold by at least 40 dB.

Bayesian Military Impulse Noise Classifier

Civilian noise complaints and damage claims have created the need for stations to monitor military impulse noise. However, the stations currently in service suffer from numerous false positive detections (due to wind noise) of impulse events and often miss many events of interest. To improve the accuracy of military impulse noise monitoring, an algorithm based upon a Bayesian classifier with inputs of conventional and custom acoustic metrics is proposed. To train and evaluate the noise classifier approximately 1,000 waveforms were field collected. The final Bayesian noise classifier used kurtosis and crest factor and, the frequency domain metrics, spectral slope and weighted square error as inputs. The EM algorithm is utilized to fit multi-Gaussian distributions to the different classes of data. In testing the classifier performed to accuracies of up to 99.6%.© 2007 ASME

Artificial Neural Network Military Impulse Noise Classifier

Civilian noise complaints and damage claims have created the need for stations to monitor the production of military impulse noise. However, these stations suffer from numerous false positive detections (due to wind noise) of impulse events and often miss many events of interest. There is also interest in identifying specific noise sources, such different types of ordinance or different types of aircraft. To improve the accuracy of military impulse noise monitoring and make and initial effort to specifically classify noise source, an algorithm based upon an artificial neural network with inputs of conventional and custom acoustic metrics was proposed. To train and evaluate the noise classifier approximately 1,000 waveforms were field collected (110 military aircraft noise, 330 military impulse noise, and 560 non-impulse noise). The final noise classifier used kurtosis and crest factor and the custom metrics spectral slope and weighted square error as inputs. The classifier was able to achieve 99.7% accuracy on the training data set and 99.4% accuracy on the validation data set.Copyright

An Investigation of the Characteristics of a Bayesian Military Impulse Noise Classifier

In extension of previous methods to identify military impulse noise in the civilian environmental noise monitoring setting by means of a set of computed scalar metrics input to artificial neural network structures, Bayesian methods are investigated to classify the same dataset. Four interesting cases are identified and analyzed: A) Maximum accuracy achieve on training data, B) Maximum overall accuracy on blind testing data, C) Maximum accuracy on testing data with zero false positive detections, D) Maximum accuracy on testing data with zero false negative rejections. The first case is used to illustrative example and the later three represent actual monitoring modes. All of the cases are compared and contrasted to illuminate respective strengths and weaknesses. Overall accuracies of up to 99.8% are observed with no false negative rejections and accuracies of up to 98.4% are also achieved with no false positive detections.Copyright

Development of artificial neural network classifier to identify military impulse noise

Civilian noise complaints from urban encroachment have curtailed military exercises that produce high levels of impulse noise. Currently, monitoring stations are in place around many military installations to monitor noise levels. However, these monitoring stations do not provide real‐time data and suffer from false positive detections. This project sought to develop algorithms with improved noise classification accuracy. Various types of military impulse noise and nonimpulse noise were measured and processed. Approximately 1000 waveforms were field collected (670 nonimpulse and 330 military impulse). Although several metrics were examined, the conventional time domain metrics of kurtosis and crest factor as well as two novel frequency domain metrics, weighted square error and spectral slope, were found to work best. The computed metric values were used to train and verify the performance of the artificial neural network (ANN), which was able to achieve 100% accuracy on training data and 100% accuracy on ...

Microphone array techniques using cross‐correlations.

Civilian noise complaints and damage claims have created a need to establish a detailed record of impulse noise generated at military training facilities. Wind noise is causing false positive impulse detections in the current noise monitoring systems. Multiple channel data methods were investigated in order to distinguish the characteristics of noise events. A microphone array was used to collect four simultaneous channels of military impulse and wind noise data. Cross‐correlation functions were then used to characterize the input waveforms. Three different analyses of microphone array data were developed. A new value, the min peak correlation coefficient, is defined from the minimum value of peaks of the cross‐correlation coefficient functions among the different channels. This value is a measure of the likelihood that a given waveform originated from a correlated noise source. The angle of incidence of the noise source is calculated using a sound source localization technique based on the geometry of th...

An image processing based neural network method of wave form classification.

In an effort to identify military impulse noise, as it relates to civilian damage and disturbance claims, several metric based approaches utilizing artificial neural networks and Bayesian classifiers have proven successful in addressing this issue. However, in the course of research, it became apparent that the noise sources to be classified, namely, various types of military impulse noise, wind noise, and aircraft noise, could be easily identified by a minimally trained observer by way of a simple visual inspection of the wave form. Additionally, since the noise classification algorithm is desired to be implemented on DSP boards with possibly limited computational resources, it may prove beneficial to avoid the computation of metrics which involve complex mathematical operations. Borrowing from proven artificial neural network techniques already proven in the field of optical character recognition, this proposed noise classifier views a captured wave form as a number of points located on a spatial grid. ...

Comparison of artificial neural network structures to identify military impulse noise

To monitor the production of military impulse noise in the area of military installation a classifier is being developed to identify military impulse events from other noise sources. In previous work [B. A. Bucci, J. S. Vipperman, J. Accoust. Soc. Am. 119, 3384 (2006)], efforts were made to identify military impulse noise from other noise sources using a multi‐layer perceptron neural network. The network used scalar input metrics (kurtosis, crest factor, spectral slope, and weighted square error) computed from recorded wave forms. In an extension to this effort, various types of neural network classifiers utilizing wavelet coefficients and cepstral coefficients as inputs are currently being tested and compared. These network types, which have proven successful in similar applications, include multi‐layer perceptrons, radial basis function network, self‐organizing maps, and recurrent networks. The networks are trained on an increased library of waveforms as compared to the previous effort. The goal of this...