تخمين و رديابي گام سيگنال هاي صوتي در شرايط غير خطي و غير گوسي

چكيده انگليسي :

Estimation and tracking of Pitch of Audio Signals in Non Linear and Non Gaussian Conditions Habib Hajimolahoseini Ha hajimolahoseini@ec iut ac ir 14th of September 2016 Department of Electrical EngineeringIsfahan University of Technology 84156 83111 Isfahan Iran1st Supervisor Dr Rassoul Amirfattahi fattahi@cc iut ac ir1st Advisor Dr Hamid Soltanian Zadeh hszadeh@ut ac ir2nd Advisor Dr Mohammad Mahdi Naghsh mm naghsh@cc iut ac irDepartment Graduate Program Coordinator Dr Mohammad Reza Taban Dep of Electrical and Computer Engineering Isfahan University of Technology 84156 83111 Isfahan Iran School of Electrical and Computer Engineering University of Tehran 14395 515 Tehran IranAbstractIn this thesis we introduce an algorithm for estimating and tracking the pitch period of audio signals usingBayesian filters For this purpose we propose Bayesian model which is robust to the non stationary variationsof the amplitude and frequency of the input signal We also employ a state space model which uses the delayedversions of the input signal to model the periodicity of non stationary audio signals This simple model allows asignificant reduction of the required number of particles for the estimation of the pitch period compared to thestate of the art particle filtering methods Moreover we propose to estimate the logarithm of the period insteadof the period itself We show that the resulting algorithm does not require prior knowledge about the initial stateand is robust to the octave error phenomenon which is a common problem in pitch period estimationmethods Our method often results in a higher time domain resolution with no perceptible compromise on thefrequency domain resolution especially for high pitched audio signals such as music Experimental resultsreveal that the proposed algorithm outperforms the state of the art pitch period detection algorithms at lowsignal to noise ratios assuming no prior knowledge about the initial conditions Key WordsPitch Period Estimation Bayesian Filtering Audio Signals State Space Model Real Time estimationIntroductionPitch period of a periodic signal is defined as the smallest positive time shift that leaves thesignal invariant Although the definition of pitch period is conceptually different from that ofthe fundamental frequency called F0 they are in a close relation so that in most of the casesF0 estimation algorithms are also called pitch period estimation methods Hess 1982 Pitchperiod or F0 estimation of periodic signals has a wide range of applications in many areas ofsignal processing including speech Jurafsky and Martin 2014 and music Muller et al 2011 In speech processing F0 variations is related to prosody and estimated in many speechrecognition systems In the case of musical signal processing F0 contributes to the musicalnotes and is estimated in automatic music transcription systems In general an audio signalmay be considered as a quasi periodic signal with time varying amplitude and pitchperiod which is often embedded in noise In this thesis we aim to estimate its time varyingpitch period Many algorithms have been proposed in the literature for pitch period estimation of audiosignals The existing methods may be categorized into two main groups parametric and non parametric algorithms The parametric algorithms introduce a model which describes thesignal being processed and therefore the problem reduces to the problem of estimating themodel parameters Bayesian Nielsen et al 2014 and sub space Christensen et al 2006 methods are two typical examples of this group On the other hand no model is used in non parametric methods and the pitch period is directly estimated by processing the signal either

حاجي ملاحسيني، حبيب

تخمين و رديابي گام سيگنال هاي صوتي در شرايط غير خطي و غير گوسي

حوزه فركانس , روش بيزي , فيلتر كالمن , خطاي اكتاو