本文選題：HE-AAC + ARM��； 參考：《西安電子科技大學(xué)》2012年碩士論文

【摘要】：HE-AAC是目前最先進(jìn)的感知音頻編解碼技術(shù),它具有多聲道、多采樣率、高壓縮比、高音質(zhì)等特點(diǎn),比AAC的效率提高30%。在許多領(lǐng)域得到廣泛的應(yīng)用,支持HE-AAC的消費(fèi)電子產(chǎn)品成為電子市場需求的熱點(diǎn)。 HE-AAC解碼器中的濾波器組包含逆向改進(jìn)余弦變換(IMDCT)、分析濾波器組和合成濾波器組三個模塊。濾波器組的基本功能是允許信號的某一部分頻率順利通過,而另外一部分頻率則受到較大抑制。濾波器組在整個音頻編解碼系統(tǒng)中,運(yùn)算量占60%以上,具有很重要的作用。隨著音頻解碼技術(shù)的不斷提高,濾波器組的研究也越來越深。 本文主要研究基于ARM的HE-AAC音頻解碼濾波器組的優(yōu)化。首先對HE-AAC音頻標(biāo)準(zhǔn)里的三個濾波器組模塊的原始算法作了分析,然后分別采用基4IFFT算法、DCT-Ⅲ算法、DCT-Ⅲ和DCT-Ⅱ的組合算法優(yōu)化IMDCT模塊、分析濾波器組以及合成濾波器組。其次,針對ARM平臺的特點(diǎn),依據(jù)定點(diǎn)實現(xiàn)原理,對濾波器組進(jìn)行定點(diǎn)轉(zhuǎn)換。最后在ARM平臺上利用NEON語法的優(yōu)勢,分別實現(xiàn)了上述三個優(yōu)化算法,并對濾波器組在寄存器分配、函數(shù)參數(shù)、循環(huán)變量、函數(shù)調(diào)用、除法運(yùn)算、乘法移位運(yùn)算、通用數(shù)據(jù)處理、加載/存儲數(shù)據(jù)等方面進(jìn)行了改進(jìn)。經(jīng)過NEON語法優(yōu)化后的濾波器組比C語言和ARMv6語法優(yōu)化的代碼大小都小,使得存儲器的使用大大減少了。并且在系統(tǒng)運(yùn)行后,通過了HE-AAC音頻標(biāo)準(zhǔn)規(guī)定的Conformance測試,提高了解碼效率。
[Abstract]:HE-AAC is the most advanced perceptual audio coding and decoding technology. It has the characteristics of multi-channel, multi-sampling rate, high compression ratio, high sound quality and so on. It is 30 times more efficient than AAC.It has been widely used in many fields, and consumer electronics supporting HE-AAC has become a hot spot in electronic market.The filter banks in HE-AAC decoder include three modules: inverse improved cosine transform (HE-AAC), filter bank and composite filter bank.The basic function of the filter bank is to allow one part of the signal to pass smoothly, while the other part is greatly suppressed.In the whole audio coding and decoding system, the filter bank accounts for more than 60% of the computation, which plays an important role.With the improvement of audio decoding technology, the research of filter banks is becoming more and more deep.This paper focuses on the optimization of HE-AAC audio decoding filter banks based on ARM.Firstly, the original algorithm of three filter banks in HE-AAC audio standard is analyzed, then the combination algorithm of DCT- 鈪，

本文編號：1751740