08523twfl.doc/006 修正日期92.10.14 玖、發明說明： 本發明是有關於一種用於音頻訊號解碼中之方法及裝 置，且特別是有關於一種用於MPEG Layer3音頻訊號解碼 中之逆修正型餘弦轉換及重疊相加之方法及裝置。 數位音頻訊號的處理在今日社會中有極廣泛的運 用。這是由於數位音頻訊號對噪音的免疫能力較類比音頻 訊號強。然而由於往往必須在極短的時間內處理龐大的資 料，且依然能維持高音質的效果，所以爲了有效解決這些 問題，便發展出許多音頻訊號壓縮標準。其中，動態影像 壓縮標準（Motion Picture Experts Group，簡稱 MPEG)以其 高壓縮率與低失真而廣受歡迎。其利用人耳對不同頻帶的 音訊有不同的敏感度，而將較不敏感的音訊給予較少的位 元，以達到壓縮的目的。 此外，爲適應不同等級的音訊品質及壓縮方法，MPEG 可再分爲第一層（Layerl)、第二層(Layer2)及第三層（Layer3) 三種處理。一般而言，愈高等級的Layer，壓縮方法愈複 雜，相對的所還原的音頻訊號失真也愈少，效果愈佳。 MPEG的編碼過程可分爲編碼器（Encoder)及解碼器 (Decoder)兩部分來硏究。在編碼器的部分中，係利用分析 次頻帶爐波器(Analysis Subband Filter Bank)而將音頻資料 加以處理成32個次頻帶（Subband)的資料。再根據模擬人 工耳聽覺效應的知覺模型(Psychoa-Coustical Model)，可以 將屬於不同頻帶的資料給予不同的位元，再經過量化處理 (Quantization)，而達到壓縮的目的。最後將資料以一定之 08523twfl.doc/006 修正日期92.10.14 資料格式包裝(Framing)，而傳送出去。 在解碼器的部份中，宛如編碼器的反動作，係先將 資料解開（Unpacking)，在經過逆量化處理（Inverse Quantization)後，最後再利用合成次頻帶濾波器（Symhesis Subband Filter Bank)，而將32個次頻帶的資料整合成原 音訊資料。 至於MPEG_II音頻編碼標準，除了提供多聲道 (Multichannel)音訊編碼，其餘基本上是與MPEG I相同的。 多聲道音訊可區分爲經由基本傳輸通道（Basic Transmission Channel)T0，Tl 傳輸的左（L)及右（R)聲道音 曰只’以及經由延伸傳転[j通道（Extended Transmission Channel)T2，T3，T4傳輸的中（C)、左環繞(LS)與右環繞(rs) 聲道音訊兩部分。MPEG-II音訊解碼需要多聲道解碼器 (Multichannel Decoder)，以重建多聲道的音頻訊號。 在 MPEG LAYER3 壓縮標準中，MPEG Layer3 (MP3) 壓縮演算法已經廣泛採用於數位廣播與多媒體的應用中。 對於數位音頻訊號壓縮而言，MP3是MPEG中最複雜但其 壓縮量也是最高的演算法。MP3之所以能達到如此高的壓 縮率，是由於第三層中利用到了逆修正型餘弦轉換 (Inverse-Modified Discrete Cosine Transform，簡稱 IMDCT) 及次頻帶編碼技術（Subband Coding)。 在硬體實現方面，MPEG之第一層及第二層的解碼器 已有許多的硏究將其裝置真正實現出來。然而MP3卻沒 有適當的裝置來實現。現今大部分的裝置設計是利用通用 08523twfl.doc/006 修正日期92.10.14 型的的數位訊號處理器(Digital Signal Processor，簡稱DSP) 來實現。此種設計乃是利用程式控制來達成。但是此種設 計需要龐大的記憶體以存放程式碼，所以加重了硬體的負 擔與面積，而使得整個系統的效能不能達到最佳化的境 地。 有鑑於此，本發明提出一種用於MPEG Layer3音頻 訊號解碼中之逆修正型餘弦轉換及重疊相加之方法及裝 置。本發明係藉由逆修正型餘弦轉換及重疊相加之快速的 演算法來將整個裝置實現出來，而使整個系統達到低花費 及高效能的需求。 爲達成上述及其他目的，本發明提出一種用於mpeg Layer3音頻訊5虎解碼中之逆修正型餘弦轉換及重疊相加之 方法。首先將壓縮音頻訊號之32個次頻帶樣本依式（1)進 行逆修正型餘弦轉換及重疊相加， 3n 4 逆修正型餘弦轉換： x(〇 = XX(^)*cos(/^) 重疊相加： 0) η Ζ{ί) = χ{ΐ) * win(i,p) ζ(~ ~ 1 ~ 0 = -χ{ΐ) * win(^ -1 - i^p) Ζ{ί) = χ{ι) * win{i,p) Ζ{η -1 ~ = χ{{) * win(n -1 - 其中，X(k)爲次頻帶樣本，ζ⑴爲經過處理後的次頻帶樣 本，當視窗類別爲0，1，3時，η等於36，而當視窗類別爲 2時，η等於I2。接著，提供動態視窗逆修正型餘弦轉換 08523twfl .doc/006 修正日期92.10.14 (DWIMDCT)模組，以動態視窗逆修正型餘弦轉換模組中 之乘加器做該逆修正型餘弦轉換的運算，並將逆修正型餘 弦轉換的運算結果儲存於動態視窗逆修正型餘弦轉換模組 中之暫存器堆中。之後，以乘加器做該重疊相加的運算， 並將重疊相加的運算結果儲存於動態視窗逆修正型餘弦轉 換緩衝記憶體中。 本發明還提出一種用於MPEG Layer3音頻訊號解碼 中之逆修正型餘弦轉換及重疊相加之裝置。此裝置包括動 態視窗逆修正型餘弦轉換模組及動態視窗逆修正型餘弦轉 換緩衝記憶體。其中，動態視窗逆修正型餘弦轉換模組包 括用以計算逆修正型餘弦轉換及重疊相加之乘加器，以及 耦接至乘加器，用以儲存逆修正型餘弦轉換的運算結果之 暫存器堆。而動態視窗逆修正型餘弦轉換緩衝記憶體係耦 接至動態視窗逆修正型餘弦轉換模組，用以儲存重疊相加 的運算結果。 綜上所述，本發明藉由逆修正型餘弦轉換及重疊相加 之快速的演算法來將整個裝置實現出來，而使整個系統達 到低花費及高效能的需求。 爲讓本發明之上述和其他目的、特徵和優點，能更加 明顯易懂，下文特舉較佳實施例，並配合所附圖示，做詳 細說明如下： 圖式簡單說明： 第1圖繪不的繪不的是根據本發明之用於mpeg Layer3音頻訊號解碼中之逆修正型餘弦轉換及重疊相加之 〇8523trwfl .doc/006 修正日期92.10.14 方法及裝置之MP3的解碼流程圖； 第2圖繪示的是根據本發明一較佳實施例之用於 MPEG Layer3音頻訊號解碼中之逆修正型餘弦轉換及重疊 相加之方法的流程圖；08523twfl.doc / 006 Revised date 92.10.14 发明 、 Explanation of the invention: The present invention relates to a method and device for decoding audio signals, and in particular to an inverse correction type for decoding MPEG Layer3 audio signals Method and device for cosine transformation and overlap addition. The processing of digital audio signals is widely used in today's society. This is because digital audio signals are more immune to noise than analog audio signals. However, because of the huge amount of data that must be processed in a very short period of time and still maintain high sound quality, many audio signal compression standards have been developed in order to effectively solve these problems. Among them, the Motion Picture Experts Group (MPEG for short) is popular for its high compression rate and low distortion. It uses the human ear to have different sensitivities to audio in different frequency bands, while giving less sensitive audio less bits to achieve the purpose of compression. In addition, in order to adapt to different levels of audio quality and compression methods, MPEG can be further divided into three layers: Layer1, Layer2, and Layer3. In general, the higher the level of the layer, the more complicated the compression method, and the less the distortion of the restored audio signal, the better the effect. The encoding process of MPEG can be divided into two parts: Encoder and Decoder. In the part of the encoder, the audio data is processed into 32 subbands by analyzing the subband filter bank. According to the Psychoa-Coustical Model that simulates the auditory effect of the artificial ear, data belonging to different frequency bands can be given to different bits, and then quantized to achieve the purpose of compression. Finally, the information was packaged in a certain 08523twfl.doc / 006 date of revision 92.10.14 (Framing) and transmitted. In the part of the decoder, like the reverse action of the encoder, the data is unpacked first, after inverse quantization, and finally the Symhesis Subband Filter Bank is used. , And integrate the 32 sub-band data into the original audio data. As for the MPEG_II audio coding standard, except that it provides multi-channel (Multichannel) audio coding, the rest are basically the same as MPEG I. Multi-channel audio can be divided into the left (L) and right (R) channel sounds transmitted through the Basic Transmission Channel T0, Tl and the extended transmission channel [jChannel (Extended Transmission Channel) T2 , T3, T4 The middle (C), left surround (LS) and right surround (rs) channel audio. MPEG-II audio decoding requires a multichannel decoder (Multichannel Decoder) to reconstruct multi-channel audio signals. In the MPEG LAYER3 compression standard, the MPEG Layer3 (MP3) compression algorithm has been widely used in digital broadcast and multimedia applications. For digital audio signal compression, MP3 is the most complex algorithm in MPEG, but its compression is also the highest. The reason why MP3 can achieve such a high compression rate is that the third layer uses Inverse-Modified Discrete Cosine Transform (IMDCT) and Subband Coding. In terms of hardware implementation, many decoders of the first layer and second layer of MPEG have been researched to realize their devices. However, MP3 has not been implemented properly. Most device designs today are implemented using a general-purpose 08523twfl.doc / 006 revision date 92.10.14 type Digital Signal Processor (DSP). This design is achieved using program control. However, this design requires a huge amount of memory to store the code, so it increases the burden and area of the hardware, so that the performance of the entire system cannot be optimized. In view of this, the present invention proposes a method and a device for inversely modified cosine conversion and overlap addition in MPEG Layer3 audio signal decoding. The present invention realizes the entire device through a fast algorithm of inverse modified cosine transformation and overlap and addition, so that the entire system achieves the requirements of low cost and high performance. In order to achieve the above and other objectives, the present invention proposes a method for inverse modified cosine conversion and overlap addition in mpeg Layer3 audio 5 tiger decoding. First, the 32 sub-band samples of the compressed audio signal are inverse modified cosine transformation and overlap addition according to formula (1), and 3n 4 inverse modified cosine transformation: x (〇 = XX (^) * cos (/ ^) overlap Addition: 0) η AZ {ί) = χ {ΐ) * win (i, p) ζ (~ ~ 1 ~ 0 = -χ {ΐ) * win (^ -1-i ^ p) ZZ {ί) = χ {ι) * win {i, p) Zn {η -1 ~ = χ {{) * win (n -1-where X (k) is the sub-band sample and ζ⑴ is the processed sub-band sample When the window type is 0, 1, 3, η is equal to 36, and when the window type is 2, η is equal to I2. Next, a dynamic window inverse correction cosine transform is provided. 08523twfl .doc / 006 Modification date 92.10.14 (DWIMDCT ) Module, the multiplier and adder in the dynamic window inverse correction cosine conversion module is used to perform the operation of the inverse correction type cosine conversion, and the operation result of the inverse correction type cosine conversion is stored in the dynamic window inverse correction type cosine conversion module Then, the overlapping and adding operation is performed by a multiplier and adder, and the operation result of the overlapping and adding is stored in a dynamic window inverse correction type cosine transformation buffer memory. The present invention also provides a to Device for inverse modified cosine conversion and overlapping addition in MPEG Layer3 audio signal decoding. This device includes a dynamic window inverse modified cosine conversion module and a dynamic window inverse modified cosine conversion buffer memory. Among them, the dynamic window inverse modified type The cosine conversion module includes a multiplier for calculating inversely modified cosine transformation and overlapping addition, and a register stack coupled to the multiplier and accumulator for storing the operation result of the inversely modified cosine transformation. And the dynamic window The inverse modified cosine conversion buffer memory system is coupled to the dynamic window inverse modified cosine conversion module, which is used to store the operation results of overlap and addition. In summary, the present invention uses inverse modified cosine conversion and overlap and addition. A fast algorithm is used to implement the entire device, so that the entire system meets the requirements of low cost and high performance. In order to make the above and other objects, features, and advantages of the present invention more obvious and understandable, the following specific implementation is preferred Examples, and the accompanying drawings, will be described in detail as follows: Brief description of the drawings: The first drawing cannot be drawn according to the present invention. Inverse modified cosine transform and overlap addition in mpeg Layer3 audio signal decoding. 8523trwfl .doc / 006 Modification date 92.10.14 Method and device MP3 decoding flow chart; Figure 2 shows a method according to the present invention. A flowchart of a method for inversely modified cosine transform and overlap and add in MPEG Layer3 audio signal decoding in a preferred embodiment;
第3圖繪不的是根據本發明一較佳實施例之用於MPEGFIG. 3 does not show a diagram for MPEG according to a preferred embodiment of the present invention.
Layer3音頻訊號解碼中之逆修正型餘弦轉換及重疊相加之 裝置圖； 第4圖繪示的是第3圖之DWIMDCT緩衝記憶體的規 劃圖；以及 第5圖繪示的是第3圖之DWIMDCT緩衝記憶體中 的記憶庫寫入與合成濾波器讀取的順序之示意圖。 重要元件標號： 10 :前處理部分 12 :後處理部分 30 : DWIMDCT 模組 32 : DWIMDCT緩衝記憶體 34 :合成濾波器模組 302 :乘加器 304 :暫存器堆 S102-sl20 : MP3的解碼流程之施行步驟 較佳實施例:Device diagram of inverse modified cosine transform and overlap and add in Layer3 audio signal decoding; Figure 4 shows the DWIMDCT buffer memory planning diagram of Figure 3; and Figure 5 shows the layout of Figure 3 Schematic diagram of the sequence of writing to the memory bank and reading from the synthesis filter in the DWIMDCT buffer memory. Important component numbers: 10: pre-processing part 12: post-processing part 30: DWIMDCT module 32: DWIMDCT buffer memory 34: synthesis filter module 302: multiplier and adder 304: register stack S102-sl20: MP3 decoding A preferred embodiment of the execution steps of the process:
本發明適用於MPEG Layer3,無論MPEG-Ι或MPEG-II 均可實施之音頻訊號解碼器設計。對於數位音頻訊號壓縮 而言，MPEG Layer3(MP3)壓縮演算法是MPEG中最複雑 08523twfl .doc/006 修正日期92.10.14 但其壓縮量也是最高的演算法。所以本發明所提供的較佳 實施例首先針對整個MP3壓縮演算法做充分的硏究，以 使資料與運算量的減少，而提出了快速演算法。接著再利 用此快速演算法將整個裝置實現出來，而使整個系統達到 低成本及高效能的需求。 請參照第1圖，其繪示的是根據本發明之用於MPEG Layer3音頻訊號解碼中之逆修正型餘弦轉換及重疊相加之 方法及裝置之MP3的解碼流程圖。整個流程可以分爲前 處理部分10及後處理部分12。前處理部分1〇首先取得位 元流並找到標頭1〇(步驟sl02)。然後進行旁註資訊的解碼 (步驟sl04)、比例因子的解碼（步驟si06)、以及賀夫曼資 料的解碼（步驟sl08)。接著將頻譜線重排(步驟sll())，並 且合倂立體聲處理（若使用時）(步驟S112)。之後，消除反 折失真（步驟S114)。前處理部分10主要負責位元流的處 理，通常這部分的運算量不大，但是過程繁瑣。所以這部 分通常可以利用有限狀態機及嵌入式的微控制器來實現， 如此便可以有效簡化整個系統的設計。近幾年來已經有許 多的硏究針對這個部分做了許多的探討。 而後處理部分12主要包含了本發明之逆修正型餘弦 轉換（Inverse-Modified Discrete Cosine Transform ’ 簡稱 IMDCT)及重疊相加（Overlap-add)的合成處理（這兩者合稱 爲動態視窗逆修正型餘弦轉換(Dynamic Window Inverse-Modified Discrete Cosine Transform，簡稱 DWIMDCT))(如 步驟si 16)，以及合成濃波器（Synthesis Filter Bank)的處理 08523twfl.doc/006 修正日期92.10.14 (如步驟S118)。而在合成濾波器處理後，會輸出脈碼調變 樣本（如步驟S120)。通常後處理部分12的運算量較大， 大約佔整個流程的80%。基於這個理由，後處理部分12 便需要一個適當的裝置來實現，才能使整個系統達到低成 本與高效能的需求。 爲使整個系統達到低成本與高效能的需求，本發明 提出了用於MPEG Layer3音頻訊號解碼中之逆修正型餘弦 轉換及重疊相加之方法（也就是IMDCT快速演算法（fast IMDCT))之一較佳實施例的流程圖，如第2圖所繪示。此 IMDCT快速演算法係將壓縮音頻訊號之32個次頻帶樣本 依式（1)進行逆修正型餘弦轉換及重疊相加， 逆修正型餘弦轉換： x{i) = ^X{k) *cos(/?A：)The invention is applicable to the design of an audio signal decoder which can be implemented in MPEG Layer 3, MPEG-I or MPEG-II. For digital audio signal compression, the MPEG Layer3 (MP3) compression algorithm is the most complex algorithm in MPEG 08523twfl .doc / 006 revision date 92.10.14, but its compression amount is also the highest algorithm. Therefore, the preferred embodiment provided by the present invention firstly conducts a thorough research on the entire MP3 compression algorithm, so as to reduce the amount of data and operation, and proposes a fast algorithm. Then use this fast algorithm to realize the entire device, so that the entire system meets the requirements of low cost and high performance. Please refer to FIG. 1, which illustrates a MP3 decoding flowchart of a method and device for inverse modified cosine conversion and overlap addition in MPEG Layer 3 audio signal decoding according to the present invention. The entire process can be divided into a pre-processing section 10 and a post-processing section 12. The pre-processing section 10 first obtains the bit stream and finds the header 10 (step sl02). Then, the marginal information is decoded (step sl04), the scale factor is decoded (step si06), and the Huffman data is decoded (step sl08). Then, the spectrum lines are rearranged (step s11 ()), and stereo processing (if used) is combined (step S112). After that, the reverse distortion is eliminated (step S114). The pre-processing part 10 is mainly responsible for processing the bit stream. Usually, this part has a small amount of calculation, but the process is complicated. So this part can usually be realized by finite state machine and embedded microcontroller, so it can effectively simplify the design of the entire system. In recent years, many studies have done a lot of discussions on this part. The post-processing part 12 mainly includes the inverse-modified discrete cosine transform (IMDCT) and the overlap-add synthesis process of the present invention (these two are collectively referred to as a dynamic window inverse-correction type Cosine Transformation (Dynamic Window Inverse-Modified Discrete Cosine Transform (DWIMDCT)) (such as step si 16), and the processing of the Synthesizer Concentrator (Synthesis Filter Bank) 08523twfl.doc / 006 Date of revision 92.10.14 (such as step S118) . After processing by the synthesis filter, pulse code modulation samples are output (as in step S120). Generally, the post-processing part 12 has a large amount of calculation, which accounts for about 80% of the entire flow. For this reason, the post-processing part 12 needs an appropriate device to realize it, so that the entire system can meet the requirements of low cost and high performance. In order to make the entire system meet the requirements of low cost and high performance, the present invention proposes a method of inverse modified cosine transformation and overlap addition in MPEG Layer3 audio signal decoding (that is, IMDCT fast IMDCT) A flowchart of a preferred embodiment is shown in FIG. 2. This IMDCT fast algorithm performs inverse modified cosine conversion and overlap addition on the 32 sub-band samples of the compressed audio signal according to formula (1). Inverse modified cosine conversion: x (i) = ^ X {k) * cos (/? A :)
k=Q ㈣十1且昏崎1 重疊相加： Z(z) = x(i) * win(i,p) (1) Z(音—1 — /) = -x(/) * (昏一 1 — /，/〇 0 < / < ^ -1 Z(/) = x(i) * win(i,p) γχ 3λΖ Z(n -1-/) = χ(ί) * win(n -1 - i,p) — < / -1 其中，X(k)爲次頻帶樣本，Z⑴爲經過處理後的次頻帶樣 本，當視窗類別爲〇，1，3時，η等於36，而當視窗類別爲 2時，η等於12。以上之式（1)指出運算量可以由η減少成 η/2，也就是DWIMDCT中的逆修正型餘弦轉換的運算量， 可以減少成一半。表一列出了原始及本發明的演算法之運 算量的比較。由表一可知，在視窗型式〇，1，3時，本發明 569550 08523twfl.doc/006 修正日期92.10.14 與原始的比率爲 0.48MOPS(Milli〇n Operation Per Second)，而在視窗型式2時，本發明與原始的比率爲 0.42MOPS。因此可以大大地降低逆修正型餘弦轉換的運 算量。 表一 功能 視窗型式 原始 本發明 比率 (MOPS) IMDCT 型式〇，1，3 2.1 1 0.48 型式2 1 0.42 0.42 接著再利用此快速演算法將整個裝置實現出來。請參 照第3圖，其繪示的是根據本發明一較佳實施例之用於 MPEG Layer3音頻訊號解碼中之逆修正型餘弦轉換及重疊 相加之裝置圖。此裝置包括DWIMDCT模組30及 DWIMDCT緩衝記憶體32 〇其中，DWIMDCT模組30主 要包括乘加器（MAC0)302，以及暫存器堆304。此裝置的 運作方式將於底下做敘述。 首先，逆修正型餘弦轉換先利用乘加器302來計算：， 最後結果將儲存於暫存器堆304中。其中暫存器堆304包 含了 18個暫存器。經過了逆修正型餘弦轉換之後，動態 視窗的重疊相加是緊接其後的運算。動態視窗的重疊相加 仍然是利用乘加器302來完成。最後結果將儲存於 DWIMDCT緩衝記憶體32中。請參照第4圖，其繪示的 是DWIMDCT緩衝記憶體32的規劃圖，由此圖可知， DWIMDCT緩衝記憶體32包括3個記憶庫(記憶庫〇、記 11 569550 08523twfl.doc/006 修正日期92.10.14 憶庫1及記憶庫2)，每個記憶庫又分爲32個次頻區塊， 而每個次頻區塊可以儲存18個樣本資料。此外，本發明 的裝置與合成濾波器模組34，能夠形成兩級的高效能管線 處理。而DWIMDCT緩衝記憶體32中之每個記憶庫內中 的樣本資料之DWIMDCT寫入與合成濾波器模組讀取的順 序，如第5圖所示。 本發明之用於MPEG Layer3音頻訊號解碼中之逆修 正型餘弦轉換及重疊相加之裝置容易與其他模組的硬體相 容，適用於超大型積體電路(VLSI)的設計。如果能結合合 成濾波器這個硬體模組，將更能提高硬體的利用率而提升 整個解碼器操作的效能。這使得MPEG Layer3解碼器可以 以ASIC來實現，而使整個系統達到低花費與高效能的需 求。 綜上所述，本發明具有如下的優點： 1.提出低成本的逆修正型餘弦轉換及重疊相加之快 速的演算法。 2·提出適用於MPEG Layer3解碼器中之逆修正型餘 弦轉換及重疊相加之裝置。k = Q ㈣10 1 and faint saki 1 overlap and add: Z (z) = x (i) * win (i, p) (1) Z (tone—1 — /) = -x (/) * (faint -1 — /, / 〇0 < / < ^ -1 Z (/) = x (i) * win (i, p) γχ 3λZ Z (n -1- /) = χ (ί) * win ( n -1-i, p) — < / -1 where X (k) is a sub-band sample and Z⑴ is a processed sub-band sample. When the window type is 0, 1, 3, η is equal to 36. When the window type is 2, η is equal to 12. The above formula (1) indicates that the calculation amount can be reduced from η to η / 2, that is, the calculation amount of the inverse modified cosine transform in DWIMDCT can be reduced to half. Table A comparison of the calculation amount of the original and the algorithm of the present invention is listed in Table 1. As can be seen from Table 1, when the window type is 0, 1, 3, the ratio of the present invention 569550 08523twfl.doc / 006 to the original date is 92.10.14 and the original ratio is 0.48MOPS (Milli ON Operation Per Second), and in the case of window type 2, the ratio of the present invention to the original is 0.42MOPS. Therefore, the calculation amount of the inverse modified cosine transformation can be greatly reduced. Table 1 Function window type original invention Ratio (MOPS) IMDCT type 0, 1, 3 2.1 1 0.48 type 2 1 0.42 0.42 Then use this fast algorithm to implement the entire device. Please refer to Figure 3, which shows the inverse used in MPEG Layer3 audio signal decoding according to a preferred embodiment of the present invention. Device diagram of modified cosine transform and overlap addition. This device includes DWIMDCT module 30 and DWIMDCT buffer memory 32. Among them, DWIMDCT module 30 mainly includes multiplier and adder (MAC0) 302, and register stack 304. This The operation of the device will be described below. First, the inverse modified cosine transform is first calculated using the multiplier and adder 302 :, and the final result will be stored in the register stack 304. The register stack 304 contains 18 temporary registers. Register. After inverse modified cosine transformation, the overlap and addition of dynamic windows is the next operation. The overlap and addition of dynamic windows is still completed by multiplier 302. The final result will be stored in DWIMDCT buffer memory Body 32. Please refer to FIG. 4, which shows a plan view of the DWIMDCT buffer memory 32. From this figure, it can be seen that the DWIMDCT buffer memory 32 includes three memory banks (memory bank 0, note 11 5 69550 08523twfl.doc / 006 Revision date 92.10.14 Memory bank 1 and memory bank 2), each memory bank is divided into 32 sub-frequency blocks, and each sub-frequency block can store 18 sample data. In addition, the device and the synthesis filter module 34 of the present invention can form a two-stage high-efficiency pipeline process. The order in which DWIMDCT writes and reads the sample data in each bank in the DWIMDCT buffer memory 32 is shown in FIG. 5. The device for reverse-corrected cosine conversion and superposition and addition in MPEG Layer3 audio signal decoding of the present invention is easily compatible with the hardware of other modules, and is suitable for the design of a very large integrated circuit (VLSI). If it can be combined with the hardware module of the synthetic filter, it will be able to improve the utilization of the hardware and improve the performance of the entire decoder operation. This allows the MPEG Layer3 decoder to be implemented in ASIC, which enables the entire system to meet the requirements of low cost and high performance. In summary, the present invention has the following advantages: 1. A low-cost inverse modified cosine transform and a fast algorithm for superposition and addition are proposed. 2. Propose an inverse-corrected cosine transform and overlap-and-add device suitable for use in MPEG Layer3 decoders.
3.本發明的裝置可以使MPEG Layer3解碼器以ASIC 來實現，而使整個系統達到低花費與高效能的需求。 雖然本發明已以較佳實施例揭露於上’然其並非用以 限定本發明，任何熟習此技藝者，在不脫離本發明之精神 和範圍內，當可作各種之更動與潤飾，因此本發明之保護 範圍當視後附之申請專利範圍所介定者爲準。 123. The device of the present invention enables the MPEG Layer3 decoder to be implemented in ASIC, so that the entire system can meet the requirements of low cost and high performance. Although the present invention has been disclosed in the preferred embodiment above, it is not intended to limit the present invention. Any person skilled in the art can make various changes and decorations without departing from the spirit and scope of the present invention. The scope of protection of the invention shall be determined by the scope of the attached patent application. 12