Abstract:
PURPOSE: A voice decoder and a method thereof for decoding a voice frame are provided to decode a voice frame which is partitioned according to a waveform interpolation decoding method. CONSTITUTION: A pre-processor generates a parameter of a current voice frame through a parameter of a previous voice frame. A decoding unit decodes a voice frame through a parameter of the current voice frame generated from the pre-processor. The pre-processor includes an REW(Rapidly Evolving Waveform) information generating unit(610), an SEW(Slowly Evolving Waveform) information generating unit(620), and a CW information generating unit(630). The CW information generating unit combines generated information. The CW information generating unit generates the CW information of the previous voice frame.
Abstract:
본 발명은 인트라 16×16 휘도 예측시에 역양자화 및 역변환의 동작 시작 시점을 앞당겨주며동작 속도를 증대시키기 위한 SAE 계산 장치 및 그를 포함하는 H.264 코딩 장치에 관한 것으로, IQIT에 필요한 양자화된 DC 성분과 역양자화된 DC 성분을 인트라 예측 동작시에 미리 계산할 수 있도록 함으로써, 역양자화 및 역변환의 동작 시점을 앞당겨 줄 수 있을 뿐 만 아니라, 역양자화 및 역변환의 동작 속도까지도 증대시켜 줄 수 있도록 한다. H.264, 인트라 예측, DC 계수, SAE 계산, 인트라 16×16 휘도 예측
Abstract:
본 발명은 VLIW 명령어 처리 장치 및 방법에 관한 것으로, 이전 주기에서 생성된 명령어 선택 조건에 따라 VLIW(Very Long Instruction Word) 명령어에 포함된 적어도 하나의 명령어 중 하나를 선택하는 명령어 선택부; 및 상기 명령어 선택부를 통해 선택된 명령어를 처리하여, 명령어 처리 결과값과 새로운 명령어 선택 조건을 생성하는 선택 명령어 실행부를 포함하여 구성되며, 이에 의하여 조건 분기 명령어의 처리 효율을 증대시키고 VLIW 아키텍쳐의 하드웨어 크기는 감소시켜 줄 수 있도록 한다. VLIW, VLIW 아키텍쳐, 조건 분기 명령어, 조건 분기, 선택적 명령어
Abstract:
본 발명은 멀티미디어 디코딩 장치에 관한 것으로서, 마이크로프로세서의 명령을 전달하는 커맨드 버스와 디코딩 처리에 필요한 데이터를 전송하는 데이터 버스를 독립된 구조로 구현하고, 상기 커맨드 버스를 통해 마이크로프로세서가 하드웨어 디코딩 코어의 각 블록으로 명령을 전송하면, 상기 명령에 따라서 하드웨어 디코딩 코어의 각 블록이 상기 데이터 버스를 통해 메모리에 저장된 데이터를 읽어와 디코딩 처리를 수행하거나, 디코딩된 데이터를 상기 메모리에 저장하도록 구현하고, PID 매칭 유닛을 통해 하드웨어적으로 선택된 채널에 대응하는 TS 패킷을 추출하여 ES 패킷까지의 파싱을 수행하도록 구현함으로써, 마이크로프로세서의 작업 부하 및 메모리 대역폭을 감소시키면서 효율성을 높일 수 있는 것이다. H.264, T-DMB, 디코딩 장치, 커맨드 버스, 데이터 버스
Abstract:
PURPOSE: A method for extracting a probability model value from a probability model table and a symbol value decoding method using the same and a symbol value decoding apparatus are provided to reduce required memory size and operation quantity, by simplifying the process of obtaining the probability model table and the index of the probability model table. CONSTITUTION: A probability model table is fractionated and reduced(110). The probability model table comprises a plurality of probability model values. Index is adjusted based on the fractionated and reduced probability model table(120). The probability model value is searched from the probability model table using the adjusted index(130). The probability model value is extracted from the probability model table.
Abstract:
PURPOSE: A network load reduction method for multi-processor system including distributed memory and a node structure thereof are provided to reduce the data access delay by reducing the traffic generated when data request is failed. CONSTITUTION: A processor(110) controls the node and processes the data. A distributed memory(120) stores the data processed by the processor. An auxiliary memory(160) stores a sharer history table. When the node requests the shared data to a first external node and receives the data from a second external node, the sharer history table stores the second external node information and the shared data information. The node includes a cache(140) which stores the data from the first external node and the distributed memory read by the processor.
Abstract:
PURPOSE: An operating apparatus including MAC(Multiplication and Accumulation) operation and DSP(Digital Signal Processor) structure and filtering method thereof are provided to reduce the resource consumption of the DSP and enhance the whole operation ability by performing the MAC operation having the two times precision. CONSTITUTION: A first and second register(110,120) stores the n-bit data. A third register(130) stores 2n- bit data. A multiplier(140) receives the data of the first register through a first input node and receives the data of the second and third register through a second input node. The multiplier multiplies the received data from the first and second input node. An ALU(Arithmetic Logic Unit)(150) receives the operation value of the multiplier through the first input node, adds the received values from the first and second input node, and transfers the added value to the third register. The operation value of the ALU is transferred to the second input node of the ALU.
Abstract:
PURPOSE: A method and device for pixel interleaving a reference image in a single bank of a frame memory, image codec system including the same are provided to remove read latency according to a row active command. CONSTITUTION: A pixel data(522) of a reference frame which is a filter output of a reconstruction image unnecessary for video processing is interleaved with a column unit of a macro block(520). The interleaved pixel data is stored as a page unit in a signal bank(510) of a frame memory. According to a preset column address and row address, a pixel data of the reference image is stored in a desired position of a frame memory.
Abstract:
PURPOSE: An image decoding device based on multiprocessor and method thereof are provided to effectively embody multimedia decoding due to limited memory resources by minimizing communication overheat between processors. CONSTITUTION: A stream parser(410) parses skip counter of an input stream and quantization parameter by dividing an input stream by heat unit. A processor(431) obtains the skip counter and a quantization parameter and a plurality of partition stream generated through the stream parser. The processor obtains decoding information of an upper processor among an adjacency processor by heat unit. The processor parallel processes decoding of plurality of dividing stream with heat unit.
Abstract:
PURPOSE: A video encoding apparatus is provided to reduce the amount of data transmitted and received between a frame memory and an image encoding apparatus. CONSTITUTION: A coarse-grain motion estimation unit(110) performs coarse-grain motion estimation based on a current MB(Macro Block) and the SW(Search Window) regions of the current MB. A fine-grain motion estimation & motion compensation unit(111) performs fine-grain motion estimation and motion compensation based on the output of the coarse-grain motion estimation unit, a previous MB, the SW region of the previous MB, and the peripheral pixels of the SW region of the previous MB. Based on the previous MB, the fine-grain motion estimation and the output of the motion compensation unit, an encoding unit performs an encoding operation for an image.