一种虚拟三维树的声音渲染方法研究

发布时间：2019-01-05 18:25

【摘要】：树木三维可视化、树木三维重建是当前国内外计算机虚拟现实技术的研究重点。目前大多数学者着眼于三维树动画的研究,在树声音生成和渲染方面的研究相对较少,通过录制真实树声音或通过专业音频制作者配音的传统方法有着匹配度不高或人工成本高昂的问题,不能满足本研究的需求。针对这种问题,本文将树分成树枝和树叶两个部分,通过分析树动画中树枝以及树叶的位置关系,采用声音建模、波表合成的方法,实验出一种虚拟三维树声音渲染的方法。本文主要贡献如下:(1)树枝声音的生成。其主要步骤如下:1.采用空气动力声中卡门涡街模型,获取树枝声音的频率。2.建立频率组获取树枝的音色。3.以共轭的形式表示频率组,并为所有的频率赋予随机相位,采用逆傅里叶变换获取树枝的波表数据。4.为树枝和风添加湍流模型,将获取的模拟风速作为波表变换的输入条件,通过波表合成生成树枝声音。5.对树枝声音生成结果添加窗函数。6.对直径相近树枝使用同一张波表,减少波表的预处理时间。通过对两棵树模型进行测试,测试结果表明,树模型1的预处理时间为120s,树模型2的预处理时间为598s。(2)树叶声音的生成。其主要步骤如下:1.通过过滤高斯白噪声生成粉红噪声。2.测试声音频率,选定2000~4000Hz段的粉红噪声作为树叶声音频率,并使用逆傅里叶变换生成树叶的波表。3.根据树叶之间相对位置关系,获取不同风速下树叶碰撞数量。4.提出一种简化的运动模型,使用一张树叶的波表模拟树叶碰撞和摩擦的过程,采用湍流模型模拟的风速作为输入条件,通过波表合成生成树叶声音。使用2棵树模型进行测试,结果表明,风速的提升会使得树叶波表的生成时间明显的增加,树模型1有285片树叶,波表生成的时间为5.9s。树模型2有6420片树叶,波表生成的时间为367s。本文使用卡门涡街和粉红噪声作为树声音模型,对树动画匹配进行了测试,实验结果表明本方法生成的树叶和树叶声音可以根据树动画中树枝和树叶位置的变化,渲染出较为匹配动画效果的树声音,并能根据观测点位置的变化,得到左右声道声效不同的树声音,满足课题的要求。
[Abstract]:Tree 3D visualization and tree 3D reconstruction are the research focus of computer virtual reality technology at home and abroad. At present, most scholars focus on the research of 3D tree animation, but the research on tree sound generation and rendering is relatively few. The traditional methods of recording real tree sound or dubbing by professional audio producers have the problem of low matching degree or high labor cost, which can not meet the needs of this study. In order to solve this problem, this paper divides the tree into two parts: tree branches and leaves. By analyzing the relationship between branches and leaves in tree animation, using the method of sound modeling and wave table synthesis, a virtual 3D tree sound rendering method is proposed. The main contributions of this paper are as follows: (1) the generation of branch sound. The main steps are as follows: 1. The frequency of branch sound was obtained by using the Carmen vortex model in aerodynamic sound. 2. Set up a frequency group to get the timbre of a branch. 3. The frequency group is expressed in conjugate form and the random phase is given to all frequencies. The wave table data of tree branches are obtained by inverse Fourier transform. 4. By adding turbulence model to tree branch and wind, the obtained simulated wind speed is used as input condition of wave table transformation, and the sound of tree branch is generated by wave table synthesis. Add window function to the result of tree branch sound generation. 6. The same Zhang Bo table is used for the similar diameter branches to reduce the pretreatment time of the wave meter. The test results of two tree models show that the pretreatment time of tree model 1 is 120 s, and that of tree model 2 is 598s. (2) the sound generation of leaves. The main steps are as follows: 1. By filtering Gao Si white noise to produce pink noise. 2. Test the sound frequency, select the pink noise of the 2000~4000Hz segment as the leaf sound frequency, and use inverse Fourier transform to generate the leaf wave table. 3. According to the relative position relationship between leaves, the number of leaf collision under different wind speed was obtained. 4. 4. A simplified motion model is proposed in which a wave table of a leaf is used to simulate the process of leaf collision and friction. The wind speed simulated by turbulence model is used as input condition and the sound of leaves is generated by wave table synthesis. Two tree models were used to test the results. The results showed that the time of leaf wave table formation was obviously increased with the increase of wind speed. The tree model 1 had 285 leaves and the time of wave table formation was 5.9 seconds. Tree model 2 has 6420 leaves and the time of wave table formation is 367s. In this paper, using Carmen vortex and pink noise as the tree sound model, the tree animation matching is tested. The experimental results show that the tree leaves and leaf sound generated by this method can be changed according to the tree branch and leaf position in the tree animation. The tree sound which matches the animation effect is rendered, and the tree sound with different sound effects can be obtained according to the change of the observation point position, which meets the requirements of the subject.
【学位授予单位】：西北农林科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.9

【相似文献】