Font Size: a A A

Three-Dimensional Sound Field Pick And Replay Based On Spherical Microphone Array

Posted on:2021-02-28Degree:DoctorType:Dissertation
Country:ChinaCandidate:H P LiuFull Text:PDF
GTID:1368330647456508Subject:Communication and Information System
Abstract/Summary:PDF Full Text Request
The high-order Ambisonics(HOA)-based spherical harmonics-related theory has become the main method for picking up and playing back three-dimensional spatial sound fields.It is combined with the theory of spherical microphone arrays to record sound fields.This three-dimensional audio format can be used for speakers and headphones Reproduced,the three-dimensional spatial sound field is no longer limited to the traditional principles of psychoacoustics and physical sound field simplification of multi-channel sound,but is further extended to accurate physical sound field reconstruction,but using HOA technology to obtain a near-perfect sound field reconstruction,The Ambisonics order and the number of playback speakers have harsh conditions and are difficult to apply in convenient scenarios.How to achieve end-to-end three-dimensional spatial sound field capture and Ambisonics(HOA)binaural playback system are less theoretically studied.This paper will combine binaural hearing and HOA technology to research the key technologies for obtaining and reproducing the accurate three-dimensional spatial sound field of the binaural in spherical acoustics,and establish the pick-up,encoding,decoding,playback,and objective evaluation of the three-dimensional spatial sound field from the spherical microphone array,meanwhile complete end-to-end system.First,compare and analyze the factors affecting the pick-up of the spherical microphone array,and focus on the improvement of the pick-up of the spherical microphone array and the improvement of the efficient spherical harmonic domain decomposition algorithm in the scenario of binaural virtual playback;the effect of binaural virtual playback also depends on the personalization of HRTF data and Spatial resolution,combining spherical harmonics and spherical wavelets to model HRTF to obtain better and accurate data;designing and engineering applications of objective quality evaluation models for the effects of 3D spatial sound field reconstruction.The main work and innovations of this paper are summarized as follows:(1)A new algorithm for sound source decomposition in the spherical harmonic domain is proposed.Aiming at the problem that the dynamic tracking accuracy requirements of the nonlinear model of the spherical harmonic domain and the complexity of the algorithm are high and cannot be used in real-time systems,the Kalman filter and compressed sensing method are used to track the position of the motion source.The idea of sparse reconstruction is introduced into dynamic tracking to further improve the tracking accuracy.At the same time,the tracking problem is converted into a DOA estimation problem when the incoming direction of the signal source changes with time.For the problem of computational complexity,the acquisition process breaks through the traditional method,transforms the signal processing to the spherical harmonic domain,and sets up a real-value model of DOA estimation based on the spherical array for the problem of computational complexity.Through sparse Bayesian learning and Kalman filtering,the orientation information of the moving target is obtained.Experimental data shows that the proposed algorithm improves the tracking accuracy,and the algorithm complexity is also greatly reduced.(2)Joint spherical harmonic function and spherical wavelet HRTF modeling.In view of the current mainstream spherical harmonic domain functions,HRTF modeling can only represent the rough structure of HRTF,and the problem of higher order is needed to represent local details of HRTF.A spherical wavelet HRTF modeling method is introduced to local details of HRTF.Modeling,using a small number of spherical wavelet analysis functions to effectively represent the local characteristics of HRTF,through the combination of spherical harmonics and spherical wavelet modeling can more fully represent the HRTF amplitude model without increasing the order.Numerical experimental results show that when the same number of parameters are used to represent the HRTF amplitude in all directions,compared with the 15th-order spherical harmonic modeling,the proposed joint modeling method has a smaller approximation error in the 7 k Hz frequency range.When compared with higher-order spherical harmonic modeling,the proposed joint modeling method produces a small approximation error in the frequency range of 20 k Hz.(3)The objective evaluation model of the three-dimensional sound field.A convenient and fast evaluation algorithm for sound spatial information is essential in the binaural playback end-to-end system.The objective test methods for current three-dimensional spatial sound field information are poorly compatible and executable.Based on the existing PEAQ framework,we propose to increase the binaural feature dimensions and improve the regression model,and establish a three-dimensional spatial sound field objective evaluation system with good compatibility and easy implementation that can reflect the real hearing situation.Through the standard data set test with subjective sound quality test results,the improved objective evaluation model supporting the three-dimensional sound field is highly correlated with the objective evaluation results.(4)End-to-end system for acquiring and replaying the sound field in three dimensions.Spherical acoustics-based three-dimensional spatial sound field acquisition and playback have research and single function system implementation in each link,but the overall system to achieve end-to-end and objective evaluation links is not much,and the existing dynamic binaural playback system It is often necessary to add additional head tracking equipment,which does not make full use of the sensors of the mobile device.This article is based on the previous research results covering picking,encoding,decoding,and replaying,and builds an experimental system for the problems of the current system,and analyzes the feasibility and deficiencies of the implementation of the above work through this system,which facilitates the verification and further progress of related work The work outlook provides the basis.
Keywords/Search Tags:Head-related transfer functions, 3D Audio Spatial Modeling, Spherical Harmonic function, Spherical wavelet function, Multivariate adaptive regression model
PDF Full Text Request
Related items