資源描述:
《mpeg-4 aac編碼技術(shù)研究及在定點(diǎn)dsp上的實(shí)現(xiàn)》由會(huì)員上傳分享,免費(fèi)在線閱讀,更多相關(guān)內(nèi)容在學(xué)術(shù)論文-天天文庫(kù)。
1、華中科技大學(xué)碩士學(xué)位論文MPEG-4AAC編碼技術(shù)研究及在定點(diǎn)DSP上的實(shí)現(xiàn)姓名:劉金申請(qǐng)學(xué)位級(jí)別:碩士專業(yè):計(jì)算機(jī)系統(tǒng)結(jié)構(gòu)指導(dǎo)教師:余勝生20070604華中科技大學(xué)碩士學(xué)位論文摘要MPEG-4AAC(AdvancedAudioCoding,高級(jí)音頻編碼)編碼標(biāo)準(zhǔn)代表了當(dāng)今最先進(jìn)的音頻編碼技術(shù),將在數(shù)字音頻廣播、家庭影院、網(wǎng)絡(luò)傳輸?shù)阮I(lǐng)域得到越來(lái)越廣泛的應(yīng)用。數(shù)字信號(hào)處理器(DigitalSignalProcessor,DSP)應(yīng)用于實(shí)時(shí)處理領(lǐng)域。定點(diǎn)DSP運(yùn)行速度快,功耗與成本低,性價(jià)比高,應(yīng)用范圍廣泛。結(jié)合二者優(yōu)勢(shì),在定點(diǎn)DSP上實(shí)現(xiàn)AAC實(shí)時(shí)編碼器具有較高的實(shí)用價(jià)值。AAC
2、算法本身復(fù)雜度高,要達(dá)到實(shí)時(shí)編碼的要求需要解決的關(guān)鍵問(wèn)題之一是提高算法效率。根據(jù)MPEG-4AAC編碼標(biāo)準(zhǔn)及對(duì)算法的基本原理的分析研究,算法優(yōu)化的關(guān)鍵點(diǎn)在于心理聲學(xué)模型、濾波器組和量化等模塊。這些模塊算法復(fù)雜度高,在運(yùn)算量和存儲(chǔ)器容量方面都有很高的要求。對(duì)心理聲學(xué)模型采用快速算法替代現(xiàn)有算法的方法,對(duì)量化模塊及濾波器組中的窗函數(shù)計(jì)算等采用查表與少量計(jì)算相結(jié)合的方法,從理論上可以有效地降低運(yùn)算復(fù)雜度。定點(diǎn)DSP上只能進(jìn)行定點(diǎn)數(shù)的運(yùn)算,所以將浮點(diǎn)AAC編碼程序轉(zhuǎn)化為定點(diǎn)運(yùn)算是在定點(diǎn)DSP上實(shí)現(xiàn)AAC實(shí)時(shí)編碼器需要解決的另一個(gè)關(guān)鍵問(wèn)題。對(duì)浮點(diǎn)運(yùn)算和定點(diǎn)運(yùn)算進(jìn)行研究,針對(duì)不同算法的特點(diǎn),
3、在定點(diǎn)化上有不同的解決方法,然后采用分模塊轉(zhuǎn)化的方式對(duì)AAC編碼程序進(jìn)行定點(diǎn)化。對(duì)優(yōu)化前后的AAC編碼程序的對(duì)比測(cè)試表明,采用的優(yōu)化方法是有效的,較大幅度提高了編碼速度,結(jié)果基本達(dá)到了實(shí)時(shí)編碼的要求;根據(jù)定點(diǎn)編碼程序和浮點(diǎn)編碼程序的信噪比測(cè)試,從理論和實(shí)踐兩個(gè)方面論證了該定點(diǎn)編碼算法可在提供高壓縮比的同時(shí)保證重建音頻信號(hào)的質(zhì)量。關(guān)鍵詞:高級(jí)音頻編碼;心理聲學(xué)模型;數(shù)字信號(hào)處理;優(yōu)化;定點(diǎn)計(jì)算I華中科技大學(xué)碩士學(xué)位論文AbstractThecodingstandardofMPEG-4AAC(AdvancedAudioCoding)representstheworld'smostad
4、vancedaudiocodingtechnology.Itwillbewidelyusedinmanydifferentareassuchasdigitalaudiobroadcast,digitalvoicestorage,hometheaterandnetworktransmission.DSP(Digitalsignalprocessor)isusedinreal-timeprocessingareas.Thefixed-pointDSPiswidelyused,runningathighspeedandwithhighcost-effectivenessconsider
5、ations.Accordingtotheiradvantages,theimplementoftheAACreal-timeencoderonfixed-pointDSPhaspracticality.TheAACalgorithmisverycomplicated.ToimprovetheencodeeffectisakeyissuetoimplementtheAACreal-timeencoder.WeresearchonMPEG-4AACencodingstandardsandthealgorithmprincipium.Thekeypointisoptimizingth
6、epsychoacousticmodelmodule,filterbanksmoduleandquantizermodule.Thesemodules’demandforcomputationandmemorycapacityishigh,andtheiralgorithmisverycomplicated.Thefastalgorithmisusedtoreplaceexistingalgorithminthepsychoacousticmodel,andthequantificationmoduleandthewindowfunctionusethemethodofcombi
7、ninglookupintableandasmallamountofcalculation.Intheory,itcaneffectivelyreducethecomputationalcomplexity.OnlythefixedpointcalculationcanbedoonFixedpointDSP.Theotherkeyissuetoimplementtheencoderonfixed-pointDSPisconvertingthefloating-pointdataf