Font Size: a A A

Research On Physically Based Fire Sound Synthesis Method

Posted on:2015-08-27Degree:MasterType:Thesis
Country:ChinaCandidate:Z J YuFull Text:PDF
GTID:2298330452959588Subject:Computer Science and Technology
Abstract/Summary:
Physically based fluid simulation is always one of the hot research topics incomputer graphics. Due to its important application value, fire simulation is the focusof recent researches. Although the previous researches brings many successes, theyare often dedicated to visual rendering, while sound synthesis of fire still remainsimmature, and related progress is limited. This thesis mainly proposes a fast methodfor fire sound synthesis. We divide fire sound signal into two parts, low frequency andmid-to high-frequency, and synthesize them separately. Then we synchronize thesetwo parts to generate the complete fire sound.First, we use a data-driven approach, exporting velocity and fuel burn rate of eachframe from fire simulation. We calculate velocity divergence integral surrounded byflame front defined by fuel burn rate. Then we use our defined FFT-Upsampling torefine velocity divergence integral, which can recover high frequency details as muchas possible. Finally, we differentiate velocity divergence integral on time to obtainsound pressure. Compared with the previous method, our model avoids explicitlyextracting flame front and calculating normal vector on each triangle of the fire front,which are both computing-intensive operations, to increase calculation speed incomplex scenes for ensuring the possibility for fast synthesis of fire sound. We use aMarching-Cube-Like method which is very conductive for GPU parallel computing tofurther acceleration.Then, we adopt wavelet analysis to relate sampling and reconstruction with scalespace. Based on this idea, we propose a fast method of synthesizing mid-and high-frequency of fire sound. We divide the signal space into a series of scale spaces. Weproduce high frequency noise of power-law spectra that matches theoretical andexperimental models of turbulent flames, and each part of which belongs to thecorresponding wavelet space. We use the high frequency to extend the frequencyresponse of our otherwise low-frequency sounds, finally obtaining the synchronizedcomplete fire sound signal.
Keywords/Search Tags:acoustic rendering, fire sound synthesis, FFT-Upsampling, wavelet analysis
Related items