( )Engineering Information Abstracts Part I 467

MA, USA. Sponsored by: IEEE Piscataway NJ USA. p 177-182CODEN: 002845Publication Year: 1997Abstract: This paper presents the results of work PlanningSystems Inc. performed in fulfillment of the requirements of a

Ž .Small Business Innovative Research SBIR Phase I feasibilityŽ .study for the Federal Highway Administration FHWA . The

feasibility of developing an acoustic sensing technique toŽreplace the current electromechanical loop detector, tape

.switch, pneumatic tube based methods of collecting vehicleaxle passages was investigated. We exploit the fact that thedominate source of road noise is the broadband acousticenergy radiated by vehicle tires by cross-correlating the out-puts of spatially separated roadside microphones. The resultsdemonstrate the feasibility of detecting vehicle passages bythe cross correlation processing of radiated tire noise. Thetechnique is based on technology developed as a consequenceof our work in underwater acoustic signal processing. In

Ž .English Author abstract 6 Refs. EI Order Number:98044175256Keywords: Intelligent vehicle highway systems; Motortransportation; Acoustic noise; Tires; Sensors; Microphones;Correlation theory; Acoustic signal processing; Highway trafficcontrol

Title: AUTOMATIC TRAFFIC MONITORING BY INTELLI-GENT SOUND DETECTION

( )Author s : Chen, S.; Sun, Z.P.; Bridge, B.Source: Proceedings of the 1997 IEEE Conference on Intelli-gent Transportation Systems, ITSC Nov 9-12 1997 Boston,MA, USA. Sponsored by: IEEE Piscataway NJ USA. p 171-176CODEN: 002845Publication Year: 1997Abstract: A novel traffic sensing technique is described whichutilizes a microphone array to detect the sound waves gener-ated by the road vehicles. The detected are then digitized andprocessed by an on-site computer using a correlation basedalgorithm, which extracts key data reflecting the road trafficconditions, e.g. the speed and density of vehicles on the road,automatically on-site. In comparison with existing traffic sen-sors, the proposed system offers lower installation and mainte-nance costs and is less intrusive to the surrounding builtenvironment. The results of theoretical analysis, computersimulation and some preliminary experiments are presented.

Ž .In English Author abstract 4 Refs. EI Order Number:98044175255Keywords: Highway traffic control; Microphones; Acousticwaves; Digital signal processing; Acoustic signal processing;Feature extraction; Computer simulation; Correlation theory;Algorithms; Sensors

Title: PROTECTION AGAINST JAMMING IN A SERVOSYSTEM OF OPTICAL RECORDERS

( )Author s : Ryabokon, Igor P.; Ataev, Viktor A.; Morozovsky,Oleg Y.Source: Proceedings of SPIE - The International Society forOptical Engineering Int. Conference on Optical Storage,

Imaging, and Transmission of Information May 14 1996 v 3055Kiev, Ukraine Sponsored by: SPIE - Int Soc for Opt Engineer-ing, Bellingham, WA USA. Society of Photo-Optical Instru-mentation Engineers Bellingham WA USA. p 259-263 ISSN:0277-786X CODEN: PSISDG ISBN: 0-8194-2470-6Publication Year: 1997Abstract: Anti-jamming processing of optoelectronic detectorsignals, ensures control system invariance to change of laserpower, dispersion of reflection from layer and presence ofinformation pits. Logarithm amplification of photodetectorcurrents stabilizes steepness of their characteristics at change

Ž .of laser power more than 10 times and factor of reflectionŽ .from surface 3 times . Besides, sections of tracing signal are

chosen, which are located in intervals between informationpits and free from recording. This measures increased focus-ing and tracking accuracy 2 - 3 times. In English 3 Refs. EIOrder Number: 97093812623Keywords: Optical recording; Optical data storage; Optoelec-tronic devices; Control systems; Servomechanisms; Photode-tectors

Title: HOW DO WE MAKE A GOOD INPUT IMAGE?( )Author s : Monroe, Stanley E. Jr.; Juday, Richard D.

Source: Proceedings of SPIE - The International Society forOptical Engineering Optical Pattern Recognition VIII Apr22-23 1997 v 3073 Orlando, FL, USA Sponsored by: SPIE - IntSoc for Opt Engineering, Bellingham, WA USA. Society ofPhoto-Optical Instrumentation Engineers Bellingham WAUSA. p 232-242 ISSN: 0277-786X CODEN: PSISDG ISBN:0-8194-2488-9Publication Year: 1997Abstract: We investigate some methods of transforming anelectronic video image prior to inserting it into an opticalcorrelator. In order that the results might be implemented asa video transfer function, we do only point operations. Weconsider the input characteristics of the input SLM and laosthe statistics of the input image. We model the correlator asnoiseless, and we evaluate the transformations by tow metrics:correlation intensity and peak-to- correlation-energy. In En-glish EI Order Number: 97093812924Keywords: Video signal processing; Image processing; Opticalcorrelation; Correlation detectors; Transfer functions; Lightmodulators; Statistical methods

Title: COMBINED OPTICAL AND NEURAL NETWORKFINGERPRINT MATCHING

( )Author s : Wilson, Charles L.; Watson, C.I.; Paek, Eung GiSource: Proceedings of SPIE - The International Society forOptical Engineering Optical Pattern Recognition VIII Apr22-23 1997 v 3073 Orlando, FL, USA Sponsored by: SPIE - IntSoc for Opt Engineering, Bellingham, WA USA. Society ofPhoto-Optical Instrumentation Engineers Bellingham WAUSA. p 373-382 ISSN: 0277-786X CODEN: PSISDG ISBN:0-8194-2488-9Publication Year: 1997Abstract: This paper presents results on direct optical match-ing of inked and real-time fingerprint images. Direct optical

( )Engineering Information Abstracts Part I468

correlations and hybrid optical neural network correlation areused in the matching system for inked fingerprints. Prelimi-nary results on optical matching of real-time fingerprints useoptical correlation. The test samples used in the inked imageexperiments are the fingerprint taken from NIST databaseSD-9. These images, in both binary and gray level forms, arestored in a VanderLugt correlator. Tests of typical crosscorrelations and auto correlation sensitivity for both binaryand 8 bit gray images are presented. When global correlationsare tested on a second inked image results are found to bestrongly influenced by plastic distortion of the finger. Whenthe correlations are used to generate features that are local-ized to parts of each fingerprint and combined using a neuralnetwork classification network and separate class-by-classmatching networks, 84.3 percent matching accuracy is ob-tained on a test set of 100,000 image pairs. Initial results withreal-time images suggest that the difficulties resulting fromfinger deformation can be avoided by combining many differ-ent distorted images when the hologram is constructed in thecorrelator. Testing this process will require analysis of 10-20second sequences of digital video. In English 22 Refs. EIOrder Number: 97093812938Keywords: Image processing; Pattern recognition; Neural net-works; Correlation detectors; Video signal processing; Holog-raphy; Database systems; Real time systems

(Title: AUTO-DESIGN OF HMMS HIDDEN MARKOV)MODELS USING PERSONAL COMPUTER AND SOUND

BLASTER( )Author s : Ma, Xiaochuan; Liu, Dong; Zhao, Rongchun

Source: Xibei Gongye Daxue XuebaorJournal of Northwest-ern Polytechnical University v 15 n 3 1997 NPU Xi’an China.p 482-486 ISSN: 1000-2758 CODEN: XGDUE2Publication Year: 1997Abstract: This paper developed the necessary software forhidden Markov models’ auto-design using PC and sound

Ž .blaster. The achievements are as follows: 1 The reduction ofŽ .noise by enhancing SNR signal-to-noise ratio was achieved

Žwith an algorithm called RTTDC random threshold time. Ž .domain clipper . 2 Automatic production of HMMs by itera-

tion after an auto-initialization process. Methods to ensurecalculation security, storage security and model structure se-curity were discussed. The software was used to build a speechrecognition system, and the test on isolated words gave anerror rate of less than 10%, which is satisfactory for most

Ž .engineering applications. In Chinese Edited author abstract4 Refs. EI Order Number: 98044172803Keywords: Speech recognition; Mathematical models; Signalprocessing; Computer applications; Computer software

Title: INSECT-BASED DETECTOR OF BOVINE MANUREODORS

( )Author s : Raj Raman, D.; Gerhardt, Reid R.Source: Proceedings of the 1997 ASAE Annual International

Ž .Meeting. Part 3 of 3 Aug 10-14 1997 v 3 Minneapolis, MN,USA. 3 ASAE St. Joseph MI USA. 15p Paper: 975025 ISSN:0145-0166 CODEN: AAEPCZ

Publication Year: 1997Abstract: We designed, constructed, and tested a single-streaminsect odormeter for the measurement of bovine manureodors. The bio-sensing element in the odormeter is a group of

Ž .six gravid face flies Musca autumnalis DeGeer in a 360-mLchamber. A microphone, mounted above the chamber, sensessounds made by the flies’ activity. The microphone signal isprocessed by an analog root-mean-square power-measuringcircuit connected to a data acquisition system. The apparatusallows electronic measurement of insect activity. Initial experi-ments demonstrate the feasibility of this system for odor

Ž .detection yesrno . Kinetic experiments show that a dose-re-sponse curve can be produced by injecting a series of increas-ingly odorous pulses into the system. As expected, the re-peatability of this technique is dependent on control of criticalbiological variables such as the flies’ odor exposure historyand reproductive state. If face fly responses can be positivelycorrelated with the responses of human panelists, then thistechnique might enable rapid, reliable and inexpensive quanti-

Ž .fication of manure odors. In English Author abstract EIOrder Number: 98044171715Keywords: Odor control; Manures; Biosensors; Microphones;Signal processing; Data acquisition

3.4. Miscellaneous

Title: ERRORS ASSOCIATED WITH THE USE OF ADAP-TIVE DIFFERENTIAL PULSE CODE MODULATION INTHE COMPRESSION OF ISOMETRIC AND DYNAMICMYO-ELECTRIC SIGNALS

( )Author s : Chan, A.D.C.; Lovely, D.F.; Hudgins, B.Source: Medical & Biological Engineering & Computing v 36n 2 Mar 1998. Peter Peregrinus Ltd Stevenage Engl. p 215-219ISSN: 0140-0118 CODEN: MBECDYPublication Year: 1998Abstract: Muscle activity produces an electrical signal termed

Ž .the myo-electric signal MES . The MES is a useful clinicaltool, used in diagnostics and rehabilitation. This signal istypically stored in 2 bytes as 12-bit data, sampled at 3 kHz,resulting in a 6 kbyte s-1 storage requirement. ProcessingMES data requires large bit manipulations and heavy memorystorage requirements. Adaptive differential pulse code modu-

Ž .lation ADPCM is a popular and successful compressiontechnique for speech. Its application to MES would reduce12-bit data to a 4-bit representation, providing a 3:1 compres-sion. As, in most practical applications, memory is organisedin bytes, the realisable compression is 4:1, as pairs of data canbe stored in a single byte. The performance of the ADPCMcompression technique, using a real-time system at 1 kHz, 2kHz and 4 kHz sampling rates, is evaluated. The data usedinclude MES from both isometric and dynamic contractions.

Ž .The percent residual difference PRD between an un-processed and processed MES is used as a performance mea-sure. Errors in computed parameters, such as median fre-quency and variance, which are usedin clinical diagnostics,and waveform features employed in prosthetic control are alsoused to evaluate the system. The results of the study demon-