1328358 九、發明說明： 【發明所屬之技術領域】 本發明是有關於一種音訊解碼器以及音訊解碼方法， 特別是有關於一種可進行平滑濾波處理的音訊解碼器以及 音訊解碼方法。 【先前技術】 數位音訊頻壓縮技術是多媒體領域中一個非常重要的 研究課題，MPEG ( Moving Pictures Experts Group )標準所 定義的音視頻壓縮技術在多媒體產品中得到廣泛應用。 MPEG標準由國際標準組織ISO的運動圖像專家組建立， 其音頻標準包括三種不同的編碼演算法，即MPEG第一 層，MPEG第二層和MPEG第三層，三種編碼演算法複雜 度不同，因而壓縮率也不相同。音頻信號通常可以根據其 中的一種演算法來進行採樣和編碼。 音訊解碼器係用來對接收到的已編碼的採樣值進行解 碼，通常包括緩衝器，解析器，重構單元以及濾波器等。 緩衝器將音訊資料由記憶體搬運到音訊解碼器，解析器對 基礎串流（elemental stream; ES)的資料串流（data stream)進 行解析，並把解析得到的資料送入重構單元，由重構單元 解碼而得到採樣值。此時的採樣值爲經過離散余弦變換 (DCT )後的頻域採樣值，因此解碼時需要對其進行反離 散餘弦轉換（Inverse Discrete Cosine Transformation : IDCT ) 來獲得數位脈衝編碼調變採樣值，最後把數位脈波編碼調 變採樣值轉換爲類比信號，即能夠爲人類所接受的聲音。 5 -般利用渡波器進行反離散餘弦轉換處理，完成音 從頻域到時域的轉換。 有時可能在資料傳輸中産生了某些錯誤，或者是在音 訊解碼器的解碼過程中出了錯誤等等，這時，往往需要採 取-些措施來掩蔽錯誤的信號。通常是在反離散餘弦轉換 後得到的脈波編碼調變樣點中插入靜音信號。但靜音信號 會使輸出的信號不連續’如^圖中的靜音信號"所示， 在時域插入的靜音信號會在邊界處産生波形的躍變，引入 面頻衝擊’使輸出的聲音變得尖銳刺耳，對人耳造成損傷。 也有的音訊解碼器是在反離散餘弦轉換後得到的脈波編碼 s周變採樣值出錯時，重復播放出錯前最後的信號。如第2 圖中的重覆信號21處所指出的，這種方法同樣會在波形的 邊界處產生躍變，並使輸出的聲音變得尖銳刺耳，令人不 悅。 爲了渡除這種高頻干擾，有的音訊解碼器在反離散餘 弦轉換變換之後、進行輸出播放之前加人U的後處 理’其中包含插入靜音信號或重放信號，並濾除這種高頻 干擾的操作。採用這種處理方式通常需要在後處理模組中 額外增加-級低通遽波器來專門處理這種平㈣除，並需 ::外配置相應的緩衝器。如第3圖所示，後處理模組中 包3 —個及閘31 ’平滑濾波器33和平滑濾波緩衝器35。 解碼益3G解析並進行㈣散餘㈣換變換，得龍波編碼 調變知樣值《當系統發現出錯時，解碼器把靜音信號36或 重放信號37設成低，靜音信號或重放錢通過及閘取 代出錯情況下的脈波編碼調變樣點38。脈波編碼調變樣點 暫時：放在平滑渡波緩衝器中’經平滑渡波器渡波後輪出 第3圖所示的音訊解妈器在時域對出錯情況樣 Π:滑遽波，雖然在—定程度上遽除了高頻干擾，:: ;曰口 了額外的濾波器和相應的緩衝器，增加 使成本無法降低。 體’ 【發明内容】 本發明-方面在於提供—種音崎碼^進行平滑據 /理，濾、除靜音信號或重放信號帶來的高頻干擾，同日; 不祐要爲平滑濾波而增加額外的硬體。 根據本發明之一實施例，音訊解碼器包括用以接收外 部基礎串流的輸人緩衝器；對基礎串流進行解碼，獲得頻 域知=值的解碼單兀，對頻域採樣值進行反離散餘弦轉換 =加® ( wnidow)處理，獲得脈波編碼調變採樣值的濾波 态，以及用以輸出脈波編碼調變採樣值的輸出緩衝器，其 中’解碼單元亦接收輸出緩衝器反饋的溢位信號向遽波器 輸入插值，濾波器則對插值和頻域採樣值進行頻域到時域 的轉換。 。解碼單元設置了插值計數器和插值單元。當輸出緩衝 器檢查到溢位狀態，則反饋一個插值信號給解碼單元。解 碼單元在插值計數器的控制下透過插值單元向濾波器輸入 插值，濾波器對此插值進行頻域到時域的轉換，以完成平 滑遽、波。 插值單7L包括一多工器，具有至少兩個輸入值一個 輸入值爲頻域採樣值，一個輸入值爲插值。多工器在插值 (δ) 計數器的控制下，對 般處於重定模式，=值進行選擇。插值計數器一 時，選擇二:^到輸出緩衝器送來的溢位信號 濾波器通道數時，重新重定始计數’·在計數值等於 碼器的頻域採樣值作爲輪入。 、禪解 排、爲32個通道的子頻帶合成滤波器，在匿流 出的頻域採樣值與插值二起：：：通時’可將解碼器輸 理，蔣相桄w 起進仃反離散餘弦轉換和加窗處 將頻域採樣值以及插值從頻域 波編碼調變採樣值。插值替㈣以仔到脈 料-個5位元的加=:如由一個5位元的暫存 法窃構成，可進行32位計數。 ^發明另—方面在提供—種音轉碼方法 除靜音信號或重放信號帶來的高頻干擾； 時不品要4平"波而增加額外的硬體。 根據本發明之另—個實施例，音訊解碼方法包括 到的外部基礎串流進行解碼，獲得頻域採樣值；選擇B 否對頻域採樣值或是插值進行反離散餘加^ 理，以輸出脈波編碼調變採棬# “ 碼铺搡制r 檢測到輸出的脈波編 馬調❻樣值的㈣㈣賴位狀㈣可 值’並使此插值係與頻域採樣值一 和加窗處理β是停止政餘弦轉換 轉換和加窗處理。 : 面在提供-種音轉碼Μ 方法，音訊解碼器包括對接收到的外部基 = 獲得頻域採樣值的解碼單元，對頻域採樣值進行== 弦轉換和加窗處理，獲得脈波編碼調變採樣值的渡波器，、 1328358 乂及用以輸出脈波編蜗調變採樣值的輸出緩衝器。 根據本發明之另—實施例，平滑濾波方法包括當輸出 緩，器的資料#減少到—溢位狀態時，發出溢位信號^ 碼單;解碼單元在溢位信號的觸發下，送插值給濾波器。 濾波器對解碼單元送入的頻域採樣值和插值一起進行反離 散餘弦轉換變換和加窗處理。 本發明在音訊解碼器的反離散餘弦轉換之前進行平滑 濾波，在頻域時對採樣值插入靜音或重放信號，複用合成 濾波器進行平滑濾波。因而不但濾除了高頻干擾，並且免 除了對額外的平滑遽波器和緩衝器的需求，節約了電路面 積，降低了成本。 【實施方式】 請參閱第4A圖與第4B圖，本發明的音訊解碼器4〇〇 包括輸入緩衝器41、解碼單元43、濾波器45和輸出緩衝 器47。 輸入緩衝器41作爲音訊解碼器400與外部的介面，接 收解複用模組（未圖示）送來的基礎串流，並送入解碼單 兀43。解碼單元43包括解析模組431和重構模組 (reconstmction) 433。解析模组431解析基礎串流，解碼 標頭資訊，確定標頭參數，爲其他的解碼模組提供控制信 號 443。重構模組 433 分析 MpEG Audi0 Layerl/Layer2 的 音框（Frame)結構，對一音框資料的位元分配（仙 allocation)進行解碼；然後進行比例因數（scalefact〇〇的 解碼、重新取採樣點、將樣點進行反量化運算；最後進行 (S) 1328358 反規格化（de-normaiize)操作。重構模組433之輸出信號 係為頻域採樣值441經由插值單元435輸入到合成子頻帶 濾波器45。輪出緩衝器47以一均勻速度送出脈波編碼調變 才木樣值貢料給後處理模組，或直接經過模數轉換後送給播 放設備，播放設備以48K/S的穩定速度進行播放。爲保持 播放設備的穩定勻速的輸出，需使輸出緩衝器中47的資料 信號保持一定的資料量大小。1328358 IX. Description of the Invention: [Technical Field] The present invention relates to an audio decoder and an audio decoding method, and more particularly to an audio decoder and an audio decoding method capable of performing smoothing processing. [Prior Art] Digital audio compression technology is a very important research topic in the multimedia field. The audio and video compression technology defined by the MPEG (Moving Pictures Experts Group) standard is widely used in multimedia products. The MPEG standard is established by the Moving Picture Experts Group of the International Organization for Standardization ISO. Its audio standard includes three different encoding algorithms, namely MPEG first layer, MPEG second layer and MPEG third layer. The three encoding algorithms have different complexity. Therefore, the compression ratio is also different. Audio signals can usually be sampled and encoded according to one of the algorithms. The audio decoder is used to decode the received encoded sample values, and generally includes buffers, parsers, reconstruction units, filters, and the like. The buffer transfers the audio data from the memory to the audio decoder, and the parser parses the data stream of the elemental stream (ES) and sends the parsed data to the reconstruction unit. The reconstruction unit decodes to obtain a sampled value. The sample value at this time is the frequency domain sample value after the discrete cosine transform (DCT), so it needs to be inverse discrete cosine transform (IDCT) to obtain the digital pulse code modulation sample value. The digital pulse code modulated sample value is converted into an analog signal, which is a sound that can be accepted by humans. 5 - Using the wave filter to perform inverse discrete cosine transform processing, complete the conversion of the sound from the frequency domain to the time domain. Sometimes there may be some errors in the data transmission, or errors in the decoding process of the audio decoder, etc. At this time, it is often necessary to take some measures to mask the wrong signal. A mute signal is usually inserted into the pulse code modulation sample obtained after the inverse discrete cosine transform. However, the mute signal will make the output signal discontinuous. As shown in the mute signal in the figure, the mute signal inserted in the time domain will produce a waveform transition at the boundary, and the surface frequency impact will be introduced to make the output sound change. Sharp and harsh, causing damage to the human ear. There are also audio decoders that use the pulse code obtained after the inverse discrete cosine transform to change the last signal before the error. As indicated by the repeated signal 21 in Figure 2, this method also produces a transition at the boundary of the waveform and makes the output sound sharp and harsh, which is unpleasant. In order to eliminate such high-frequency interference, some audio decoders add post-processing of U after the inverse discrete cosine transform and before the output is played, which includes inserting a mute signal or a replay signal, and filtering out the high frequency. Interference operation. This type of processing usually requires an additional-stage low-pass chopper in the post-processing module to specifically handle this flat (four) divide, and the corresponding buffer needs to be configured externally. As shown in Fig. 3, the post-processing module includes a block 3 and a gate 31' smoothing filter 33 and a smoothing filter buffer 35. The decoding benefits 3G analysis and performs (4) the residual (four) conversion transformation, and the dragon wave code modulation modulation sample value "when the system finds an error, the decoder sets the mute signal 36 or the playback signal 37 to low, mute signal or replay money. The pulse code modulation modulation point 38 in the case of an error is replaced by a gate. The pulse code modulation modulation sample is temporarily placed in the smooth wave buffer. After the smoothing wave is crossed, the audio device shown in Fig. 3 is in the time domain for the error situation: the slippery wave, although in- To a certain extent, high frequency interference is eliminated, :: ; additional filters and corresponding buffers are added, so that the cost cannot be reduced. [Invention] The present invention is directed to providing a high-quality interference caused by smoothing data, filtering, removing a mute signal or a playback signal, the same day; Extra hardware. According to an embodiment of the present invention, an audio decoder includes an input buffer for receiving an external base stream; decoding the base stream to obtain a decoding unit of a frequency domain known value, and countering the frequency domain sample value Discrete cosine transform = plus (wnidow) processing, obtaining a filtered state of the pulse code modulated sample value, and an output buffer for outputting the pulse code modulated sample value, wherein the 'decoding unit also receives the output buffer feedback The overflow signal is interpolated to the chopper input, and the filter performs frequency domain to time domain conversion on the interpolated and frequency domain samples. . The decoding unit sets an interpolation counter and an interpolation unit. When the output buffer checks for an overflow condition, an interpolation signal is fed back to the decoding unit. The decoding unit inputs the interpolation to the filter through the interpolation unit under the control of the interpolation counter, and the filter performs frequency domain to time domain conversion on the interpolation to complete the smooth 遽 and wave. The interpolation list 7L includes a multiplexer having at least two input values, one input value being a frequency domain sample value, and one input value being an interpolation value. The multiplexer is normally in the re-mode and = value under the control of the interpolation (δ) counter. When the interpolation counter is selected, the selection of two: ^ to the overflow signal sent by the output buffer, the number of filter channels, re-reset the initial count '· the count value is equal to the frequency domain sample value of the encoder as a round. The meditation solution is a sub-band synthesis filter of 32 channels. When the sampled value in the frequency domain of the outflow is interpolated and the interpolation is performed: :: When the time is passed, the decoder can be processed, and Jiang Xiangyu starts the inverse discrete cosine transform. And windowing, the frequency domain sampled values and the interpolation values are modulated from the frequency domain wave coded samples. Interpolation (4) to the pulse - a 5-bit addition =: If a 5-bit temporary storage method, can be 32-bit count. ^Inventing another aspect - providing a method of transcoding, except for the high frequency interference caused by the mute signal or the replay signal; when the product is not required, the additional hardware is added. According to another embodiment of the present invention, the audio decoding method includes: decoding the external base stream to obtain a frequency domain sample value; and selecting B to perform inverse discrete addition on the frequency domain sample value or the interpolation to output Pulse code modulation modulation “ “ 码 码 码 检测 检测 检测 检测 检测 检测 检测 检测 检测 检测 检测 检测 检测 检测 检测 检测 检测 检测 检测 检测 检测 检测 检测 检测 检测 码 码 码 码 码 码 码 码 “ 码 “ 码 “ 码 “ “ “ “ “ “ “ “ “ “ “ “ β is to stop the political cosine transform conversion and windowing processing. : The surface is provided with a sound transcoding method. The audio decoder includes a decoding unit that receives the external base = obtains the frequency domain sample value, and performs the frequency domain sampling value. == Chord conversion and windowing, obtaining a pulse coded modulated sampled waver, 1328358 乂 and an output buffer for outputting a pulse wave tuned sampled value. According to another embodiment of the present invention, The smoothing filtering method includes: when the output data is reduced to the -over state, the overflow signal is sent; the decoding unit sends the interpolation value to the filter under the trigger of the overflow signal. The filter sends the signal to the decoding unit. Frequency domain sampled value The interpolation performs inverse discrete cosine transform transformation and windowing processing. The invention performs smoothing filtering before the inverse discrete cosine transform of the audio decoder, inserts a mute or playback signal into the sampled value in the frequency domain, and multiplexes the synthesis filter to smooth Filtering, thus not only filtering out high-frequency interference, but also eliminating the need for additional smoothing choppers and buffers, saving circuit area and reducing cost. [Embodiment] Please refer to Figures 4A and 4B. The audio decoder 4 of the invention comprises an input buffer 41, a decoding unit 43, a filter 45 and an output buffer 47. The input buffer 41 serves as an interface between the audio decoder 400 and the external interface, and receives the demultiplexing module (not shown). The basic stream sent is sent to the decoding unit 43. The decoding unit 43 includes an analysis module 431 and a reconstruction module 433. The analysis module 431 parses the basic stream, decodes the header information, and determines The header parameter provides a control signal 443 for other decoding modules. The reconstruction module 433 analyzes the frame structure of the MpEG Audi0 Layerl/Layer2. The bit allocation (sinallocation) is decoded; then the scaling factor (the decoding of the scalefact〇〇, the re-fetching of the sampling points, the inverse quantization of the samples), and finally the (S) 1328358 de-normaiize operation The output signal of the reconstruction module 433 is the frequency domain sample value 441 input to the synthesis sub-band filter 45 via the interpolation unit 435. The wheel-out buffer 47 sends the pulse code modulation at a uniform speed to the wood sample value. The post-processing module is sent to the playback device directly after analog-to-digital conversion, and the playback device plays at a stable speed of 48K/S. In order to maintain a stable and uniform output of the playback device, the data signal of the output buffer 47 is required. Keep a certain amount of data.
請參閱第5圖，輸出緩衝器47中設置了上溢指標5ι 和下溢指標52,分別相應於溢位信號48中的上溢信號 (almost_full)以及下溢信號（alm〇st—empty )。發錯信號、 影音不同步、解碼錯誤等可能出現的問題由cpu來控制， 當CPU發出控制信號中止了出錯單元的操作，可能會引起 CPU的回應速度出現問題。也有其他情況可能引起咖的 回應速度出現問題。當匯流排、處理器等資源的回應速度 出現問題時，冑出緩衝器中的資料可能出現溢位狀態，此Referring to FIG. 5, an overflow buffer 51 and an underflow indicator 52 are set in the output buffer 47, corresponding to the overflow signal (almost_full) and the underflow signal (alm〇st-empty) in the overflow signal 48, respectively. Possible problems such as error signal, video and audio synchronization, decoding error, etc. are controlled by the CPU. When the CPU issues a control signal to stop the operation of the error unit, it may cause a problem with the CPU response speed. There are other situations that may cause problems with the response speed of the coffee. When there is a problem with the response speed of resources such as bus, processor, etc., the data in the buffer may overflow.
時，輸出緩衝器會把這溢位信號48反饋給解碼單元Μ。解 碼單元接收到溢位信號48後，若發現溢位信號补為一下 溢信號，則選擇插值傳送給滤波器45，並㈣波器Μ對此 插值進行反離散餘弦轉換和加窗處理；若發現溢位信號48 :-上溢信號，則使遽波器停止進行反離散餘弦轉換和加 窗處理。輸入緩衝器41和輸出緩衝器47可以是先進先出 緩衝器（FIFO)。 在本發明—實施财m 45例如採用_種子頻帶 合成滤波器。子頻帶編碼是利用多相正交子頻帶分㈣波 器組把信號頻帶分割成若干子頻帶1後對各子頻帶㈣ 1328358 採用單獨的編碼方法編碼並輸 帶内的-個對應樣點作爲—個採樣电，、=碼時，母個子頻 器包含32個通道，-個採樣組辑子頻帶合成渡波 子頻帶合成遽波器對子頻帶編：二32 _點。多相正交 =行反離散餘弦轉換和加窗處:逆= =The output buffer will feed back the overflow signal 48 to the decoding unit. After the decoding unit receives the overflow signal 48, if it finds that the overflow signal is added to the overflow signal, the interpolation is selected and transmitted to the filter 45, and the (four) wave device performs inverse discrete cosine conversion and windowing on the interpolation; The overflow signal 48:-overflow signal causes the chopper to stop performing inverse discrete cosine conversion and windowing. The input buffer 41 and the output buffer 47 may be a first in first out buffer (FIFO). In the present invention - the implementation of the m 45 is, for example, a _ seed band synthesis filter. The sub-band coding uses the multi-phase orthogonal sub-band sub-four-wave group to divide the signal band into several sub-bands 1 and then encodes each sub-band (4) 1328358 with a separate coding method and a corresponding sample within the transmission band as - For the sampling power, == code, the mother sub-frequency device contains 32 channels, and the sampling group is sub-band synthesis sub-band sub-band synthesis chopper pair sub-band coding: two 32 _ points. Multiphase Orthogonal = Line Inverse Discrete Cosine Transform and Windowing: Inverse ==
器共川級，包含32組正交子子頻帶合成遽波 于頻f /慮波器，進行反離韵_絡 和加窗處理。反離散餘弦轉換將解碼單元送出 ==頻域(8域)轉換到時域(τ域)，得到脈The device has a total of 32 sets of orthogonal sub-subbands synthesized by chopping on the frequency f / filter, and performs inverse symmetry and windowing. The inverse discrete cosine transform sends the decoding unit out == frequency domain (8 domain) to time domain (τ domain), get pulse
如第4B圖所示’解碼單元43中設置了插值計數器切 口插=早π 435。輸出緩衝器檢查到溢位狀態，則反饋_個 溢位信號48(almost_empty)給解碼單元43。若此溢位作號 48為-下溢信號，解碼單元43在插值計數器437所輸出的^ 控制信f虎439下經由插值單& 435向滤波器45輸入插值， 遽波器45對解碼單％ 43所送入之插值/頻域採樣值445進 行頻域到時域的轉換，從而完成平㈣波；若此溢位信號 48為上溢#號，則濾波器止進行頻域到時域的轉換。 5月參閱第6圖，依照本發明之一實施例，插值單元435 之貫鉍方式包括多工器61，具有至少兩個輸入值，—個 輸入值爲頻域採樣值441，另一個輸入值爲插值611。多工 盗61在插值計數器437的控制下，對頻域採樣值441與插 值611進行選擇。插值計數器一般處於重定模式，此時插 值單元435挑選頻域採樣值441作爲輸入。當接收到輸出 11 :…7送來的溢位信號48時’若此溢位信號48為一下 =戒，則插值計數器437控制插值單元435選擇插值川 作為輸入，插值計數器437並同時開始計數；在計數值等 於慮波器寬度時’插值計數器重新設定，多工器則重新選 擇頻域採樣值作爲輸入。 插值單元435内之插值611可以是一般的常數值或 零:在-種實施方式中，當解解元43收到代表下溢信號 的j位信號48時’插值單元435將數值為零的插值送給滤 波器45。根據子頻帶濾波器的特點，所須插入之零值數目 與渡波器之通道數目相同，在此即32個零值。因此，插值 計數器可由一個5位的暫存器和一個5位元的加法器構成 (未顯示於圖中），來進行32位計數。 ^在本發明的另一個實施例十，子頻帶合成濾波器例如 爲1024級，包含32個通道，以32個子頻帶樣點作爲一個 採樣組，每個子頻帶包括32個點。當解碼單元收到代表下 溢信號的溢位信號48時，同樣也是插入與濾波器通道數相 同寬度的零值，即32個零值。插入的零值個數和插值計數 器的β十數位數以及濾波器通道數一致，在改變濾波器級數 與寬度的情況下，均可作相應變化。 請參閱第7圖，其係繪示本發明一實施例之音訊解碼 器在收到代表下溢信號的溢位信號時，插入插值的波形 圖。匯流排、處理器等資源的回應速度出現問題時，可能 引起輸出緩衝器出現溢位狀態。解碼單元43根據輸出緩衝 器的溢位信號48決定是否送插值給濾波器。插值產生的波 形如圖中兩條虛線之間的波形71。當插值經過濾波器渡 12 1328358 2出之脈波編碼調變波形得以連續，因此消除 邊界處産生的波形躍變。 這裏的;慮波器爲進行反離散餘弦轉換變換的子頻帶合 成濾波器’爲音訊解碼所需之元件。當匯流排、處理器； 貢源的回應速度出現問題時，需要在反離散餘弦轉換之前 插入零值使传m能夠將零值以及解碼單元送入的 域採樣值—起進行反離散餘_換和加窗處理，使之從頻 ，轉換ί丨時域m插人的零值*會造成波形的躍 餐’因此不需要額外的低通據波器進行平滑遽波。 在本發明的一實施例中，在匯流排、處理器等資源的 回應速度出現問題時，解碼單元根據前面的信號預測得到 插值，並把預測得到的插值送給驗器。如果上—音框解 媽正蜂’解碼早π可以將上—音框的解碼結果作爲插值， 重復輸出送給濾波器。 請參閱第8圖，其係繪示本發明一實施例之音訊解媽 方法流程圖。解碼步驟S8卜對接收到的外部基礎串流進行 解碼’獲仔頻域採樣值。渡波步驟如’可以在—下溢信號 f生時選擇對頻域採樣值或插值進行反離散餘弦轉換和加 窗處理’輸出脈波編碼調變採樣值（在未提供插值的情況 下δ玄步驟僅對頻域採樣值進行反離散餘弦轉換和加窗處 理h溢位狀態判斷步驟S83 ’檢測並判斷輸出的脈波編喝 調變採樣值的資料量是否達到溢位狀態，如否，則返回濾 波步驟S82,繼續對賴採樣值進行反離散触轉換和加/窗' 處理’輸出脈波編碼調變採樣值；若達—溢位狀態，且溢 位^號為一下溢信號，則執行插值提供步驟S84來提供插 13 (δ) 值，並返回濾波步驟S82,此時對插值進行反離散餘弦轉換 和加®處理並輸出脈波編碼調變採樣值。 凊參閱第9圖，其係繪示本發明一實施例之音訊解碼 器的平滑濾波方法流程圖。結合第4圖和第9圖，音訊解 碼益包括對接收到的外部基礎串流進行解碼獲得頻域採樣 2的解碼單元43，對頻域採樣值進行反離散齡轉換和加 窗處理’獲得脈波編碼調變採樣值的渡波器45，以及用以 輸出脈波編碼調變採樣值的輸出緩衝器47。平滑濾、波方法 包括：在溢位狀態判斷步驟S91巾，當匯流排、處理器等 資源的回應速度出現問題，檢測到輸出緩衝器的資料 少到-，位狀態時，會發出一代表下溢信號的溢位信號私 :解碼早το 43。插值輸出步驟S93，解碼單元在溢位作 號48的觸發下，送插值仏嗆 。 、唐丄泛播值…慮波益45。在濾、波步驟S95中’ 濾波态45對解碼單元43送入的頻域採 離散餘弦轉換和加窗處理。料_值和插值進行反 雖然本發明已以.實施例揭露如上，然 本發明，任何在此發明所屬技術領域卜且、、甫以限疋 在不脫離本發明之精神和範圍内，t 知識者， 飾，因此本發明之保護範圍當視後 =更動與潤 定者為準。 申明專利範圍所界 【圖式簡單說明】 為讓本發明之上述和其他目的 ^ 能更明顯易懂，所附圖式：詳細說明：徵、優點與實施例 第〗圖係繪示傳統音訊解：严 14出錯情况下插入靜音 1328358 信號的波形圖。 第2圖係繪示傳統音訊解碼 信號的波形圖。 弟圖係、·曰示傳統平滑滤波音訊解石馬器結構圖。 第4Α圖係繪示本發明一實施例 器的結構圖示意圖。 千-濾波“fl解碼 第4B圖係繪示第4A圖所示音訊解碼 元的結構示意圖。As shown in Fig. 4B, the interpolation counter is inserted in the decoding unit 43 = early π 435. When the output buffer checks for an overflow state, then an overflow signal 48 (almost_empty) is fed back to the decoding unit 43. If the overflow number 48 is an underflow signal, the decoding unit 43 inputs the interpolation to the filter 45 via the interpolation list & 435 under the control signal f 439 outputted by the interpolation counter 437, and the chopper 45 pairs the decoding list. The interpolation/frequency domain sample value 445 sent by % 43 performs frequency domain to time domain conversion, thereby completing the flat (four) wave; if the overflow signal 48 is the overflow # number, the filter stops the frequency domain to the time domain. Conversion. Referring to FIG. 6 in May, in accordance with an embodiment of the present invention, the interpolation unit 435 includes a multiplexer 61 having at least two input values, one input value being a frequency domain sample value 441, and the other input value. Interpolated 611. The multiplexer 61 selects the frequency domain sample value 441 and the interpolation value 611 under the control of the interpolation counter 437. The interpolation counter is typically in re-assertion mode, at which point interpolation unit 435 selects frequency domain sample value 441 as input. When the overflow signal 48 sent from the output 11:...7 is received, 'If the overflow signal 48 is next = or, the interpolation counter 437 controls the interpolation unit 435 to select the interpolation as an input, and the interpolation counter 437 starts counting simultaneously; When the count value is equal to the filter width, the interpolation counter is reset and the multiplexer reselects the frequency domain sample value as an input. The interpolation 611 in the interpolation unit 435 may be a general constant value or zero: in an embodiment, when the solution element 43 receives the j-bit signal 48 representing the underflow signal, the interpolation unit 435 interpolates the value to zero. It is sent to the filter 45. According to the characteristics of the subband filter, the number of zero values to be inserted is the same as the number of channels of the ferrator, here 32 zero values. Therefore, the interpolation counter can be composed of a 5-bit register and a 5-bit adder (not shown) for 32-bit counting. In another embodiment of the present invention, the subband synthesis filter is, for example, 1024 levels, and includes 32 channels, with 32 subband samples as one sample group, and each subband includes 32 points. When the decoding unit receives the overflow signal 48 representing the underflow signal, it also inserts a zero value of the same width as the number of filter channels, i.e., 32 zero values. The number of zeros inserted is the same as the β-digit number of the interpolation counter and the number of filter channels. When the number of filters and the width of the filter are changed, the corresponding changes can be made. Please refer to FIG. 7 , which is a waveform diagram of an interpolation inserted by an audio decoder according to an embodiment of the present invention when an overflow signal representing an underflow signal is received. When there is a problem with the response speed of resources such as bus, processor, etc., the output buffer may overflow. Decoding unit 43 determines whether to interpolate the value to the filter based on the overflow signal 48 of the output buffer. The waveform produced by the interpolation is shown as the waveform 71 between the two broken lines in the figure. When the interpolation passes through the filter wave 12 1328358 2, the pulse code modulation waveform is continuous, thus eliminating the waveform transition generated at the boundary. Here, the filter is a sub-band synthesis filter for performing inverse discrete cosine transform conversion, which is an element required for audio decoding. When there is a problem with the response speed of the bus, processor, and tribute, it is necessary to insert a zero value before the inverse discrete cosine transform so that the m can input the zero value and the domain sample value sent by the decoding unit - the inverse discrete _ And window processing, so that it is converted from the frequency, the zero value * inserted in the time domain m will cause the waveform of the leap 'so does not require additional low-pass data to smooth chopping. In an embodiment of the invention, when there is a problem with the response speed of the resources such as the bus bar and the processor, the decoding unit obtains the interpolation according to the previous signal prediction, and sends the predicted interpolation to the detector. If the upper-sound box is decoded, the mother is bee's decoding early π, and the decoding result of the upper-sound box can be used as interpolation, and the repeated output is sent to the filter. Please refer to FIG. 8 , which is a flow chart of an audio decoding method according to an embodiment of the present invention. The decoding step S8 decodes the received external base stream to obtain the frequency domain sample value. The wave step such as 'can be used to select the frequency domain sample value or the interpolation for the inverse discrete cosine transform and the windowing process when the underflow signal f is generated'. The output pulse code modulation sample value (the δ step is not provided if the interpolation is not provided) Performing inverse discrete cosine transform and windowing processing only on the frequency domain sampled value. The overflow state determination step S83 'detects and determines whether the data amount of the output pulse wave brewing modulated sample value reaches the overflow state, and if not, returns Filtering step S82, continuing to perform inverse discrete touch conversion and add/window 'processing' output pulse code modulation modulated sample values for the Lai sample value; if the overflow state is overflowed, and the overflow ^ number is a spill signal, interpolation is performed. Step S84 is provided to provide the interpolation 13 (δ) value, and the filtering step S82 is returned. At this time, the inverse discrete cosine transform and the addition processing are performed on the interpolation, and the pulse code modulated sample value is output. 第 See Fig. 9, which is drawn A flowchart of a smoothing filtering method of an audio decoder according to an embodiment of the present invention. In combination with FIG. 4 and FIG. 9, the audio decoding benefit includes decoding the received external basic stream to obtain a decoding sequence of the frequency domain sampling 2. 43. Performing anti-discrete conversion and windowing processing on the frequency domain sample values', obtaining a pulse wave coded modulated sample value of the wave 45, and outputting a buffer buffer for outputting the pulse code modulated sample value 47. Smoothing, The wave method includes: in the overflow state judging step S91, when there is a problem in the response speed of the resources such as the bus bar and the processor, when the data of the output buffer is detected to be less than -, the bit state is issued, and a signal representing the underflow signal is issued. The overflow signal is private: decoding early το 43. Interpolation output step S93, the decoding unit sends the interpolation value 触发 under the trigger of the overflow position 48. The Tang 丄 panning value... the wave benefit 45. In the filtering, wave step S95 In the filter state 45, the frequency domain is subjected to discrete cosine transform and windowing processing. The material_value and the interpolation are reversed. Although the present invention has been disclosed in the above embodiments, the present invention, any of the inventions herein The technical field of the present invention is not limited to the spirit and scope of the present invention, and therefore, the scope of protection of the present invention is subject to the change of the scope of the invention. 【figure BRIEF DESCRIPTION OF THE DRAWINGS In order to make the above and other objects of the present invention more obvious and easy to understand, the drawings: detailed description: signs, advantages and embodiments of the figure show the traditional audio solution: severe 14 error insertion 1328358 Waveform diagram of the signal. Fig. 2 is a waveform diagram of the conventional audio decoding signal. The figure shows the structure of the conventional smoothing filter and the device of the stone device. The fourth diagram shows an embodiment of the present invention. Schematic diagram of the structure diagram. Thousand-filtering "Fl decoding" Figure 4B shows the structure of the audio decoding element shown in Figure 4A.
第5圖係繪示本發明一實施例之音訊解碼器的輸出緩 衝器的原理示意圖。 第6圖係繪示本發明一實施例中的解碼單元的插值單 元結構示意圖。 第7圖係繪示本發明的音訊解碼器插入插值後的波形 圖。 第8圖係繪示本發明一實施例之音訊解碼方法流程圖。Figure 5 is a schematic diagram showing the principle of an output buffer of an audio decoder according to an embodiment of the present invention. Figure 6 is a block diagram showing the structure of an interpolation unit of a decoding unit in an embodiment of the present invention. Figure 7 is a waveform diagram showing the insertion of the interpolated audio decoder of the present invention. FIG. 8 is a flow chart showing an audio decoding method according to an embodiment of the present invention.
裔在出錯情況下插入重放 器中一個解碼單 第9圖係繪示本發明一實施例之音訊解碼器中平滑漁 波方法的流程圖。 脈波編碼調變樣點 400 :音訊解碼器 【主要元件符號說明】 11 :靜音信號 3〇 :解碼器 33 :平滑濾波器 36 :靜音信號 38 : 41 :輸入缓衝器 21 :重覆信號 3 1 :及閘 3 5 :平滑濾波緩衝器 3 7 :重放信號 15 1328358 43 :解碼單元 431 :解析模組 433 :重構模組 435 :插值單元 437 :插值計數器 439 :控制信號 441 :頻域採樣值 443 :控制信號 445 :插值/頻域採樣值 45 :濾波器 47 ： 輸出緩衝器 48 :溢位信號 51 : 上溢指標 52 :下溢指標 61 ： 多工器 611 :插值 71 :波形 s81 :解碼步驟 s82 :濾波步驟 s83 :溢位狀態判斷步驟 s84 :插值提供步驟 s91 :溢位狀態判斷步驟 s93 :插值輸出步驟 s95 :濾波步驟 (δ) 16Inserting a Decoding List in a Reproducer in the Case of Errors FIG. 9 is a flow chart showing a method of smoothing a fish wave in an audio decoder according to an embodiment of the present invention. Pulse code modulation sample 400: audio decoder [main component symbol description] 11: mute signal 3: decoder 33: smoothing filter 36: mute signal 38: 41: input buffer 21: repeated signal 3 1 : and gate 3 5 : smoothing filter buffer 3 7 : playback signal 15 1328358 43 : decoding unit 431 : analysis module 433 : reconstruction module 435 : interpolation unit 437 : interpolation counter 439 : control signal 441 : frequency domain sampling Value 443: Control signal 445: Interpolation/frequency domain sample value 45: Filter 47: Output buffer 48: Overflow signal 51: Overflow indicator 52: Underflow indicator 61: Multiplexer 611: Interpolation 71: Waveform s81: Decoding step s82: filtering step s83: overflow state judging step s84: interpolation providing step s91: overflow state judging step s93: interpolating output step s95: filtering step (δ) 16