كنترل فعال نويز بر مبناي شاخص‌هاي ادراكي شنوايي (سايكوآكوستيك)

چكيده انگليسي :

Active noise control systems based on hearing psychoacoustic metrics Seyed Mehdi Pourseiedrezaei s pourseiedrezaei@pa iut ac ir Date of Submission 2019 08 16 Faculty of Mechanical Engineering Pardis of Isfahan University of Technology Isfahan 84156 83111 Iran Degree PhD Language FarsiSupervisor Dr Mehdi Keshmiri Dr Ali LoghmaniAbstractThe active noise control ANC is being increasingly regarded in modern societies ANC is a risky andcomplex process inside closed enclosures e g inside the cabin of a car With the ever increasingdevelopments achieved in the transportation technologies during the recent past and considering thedemand for long travels and the importance of the passenger comfort during the trip the subject matter ofnoise control inside vehicles has become even more important In addition reduction of the noise generatedinside medical equipment e g infant incubator in the neonatal intensive care unit NICU has gained aparticular deal of attention recently Dealing with the broadband noise in such environments represents aserious challenge encountered by the ANC methods Only a few of the proposed techniques so far haveconsidered human auditory perception as the criterion for designing noise control systems The presentdissertation aims at modeling and developing sound quality SQ assessment and ANC methods based onthe human auditory characteristics and psychoacoustic analysis of the residual low frequency noise inclosed spaces Two sound quality evaluation SQE systems are herein proposed for non stationary broadband sounds Firstly in order to determine the objective psychoacoustic metrics a model based on multi level nodeselection MLNS of wavelet packet transform WPT combined with an optimized artificial neural network ANN is presented The sound signal is decomposed into 24 critical bands according to the human auditorycharacteristics using the MLNS WPT method The features matrix corresponding to each of the sub signalsat the selected nodes of the wavelet packet tree is fed to the developed ANN model and the objectiveparameters of the sound including the loudness sharpness roughness and tonality appear at the networkoutput In order to achieve the optimal ANN model we use the genetic algorithm GA particle swarmoptimization PSO algorithm and imperialist competitive algorithm ICA to optimize the weights andthresholds of the ANN Continuing with the work in order to lower the computational burden whileenhancing the accuracy of the SQE model a new model so called OAWT ANN is presented by combiningthe optimal analytic wavelet transform OAWT with the ANN Results of computer assisted simulationsshow that focusing on the applicability the proposed methodology can reduce the computation time whileenhancing the estimation accuracy compared to conventional SQ assessment models For the active noise control three new algorithms are developed for active sound quality control ASQC The proposed methods rely on the widespread application of active noise equalization ANE systemswhich can improve SQ of broadband disturbances by reshaping the amplitude and spectrum of the noise Inthe first control system an auditory filter designed based on the specific loudness index is integrated withthe ANC system constructed based on the filtered error least mean square FeLMS structure Due to theslow convergence rate of the first control system especially for non stationary noise a new algorithm isdeveloped based on the broadband ASQC structure which can cancel out the annoying portion of theresidual noise by applying the auditory filter that is designed based on the specific loudness Finally anASQC system is developed based on the pleasantness index wherein online empirical mode decomposition

پورسيدرضايي، مهدي

كنترل فعال نويز بر مبناي شاخص‌هاي ادراكي شنوايي (سايكوآكوستيك)

كنترل فعال كيفيت صدا , مدل‌هاي ارزيابي كيفيت صدا , شاخص‌هاي سايكوآكوستيك , نويز باند پهن , نويز غير ايستان