Abstract: A coding apparatus and a coding method are proposed. The coding apparatus includes a memory and a processor. The processor is configured to obtain a feature map, perform lossy compression on the feature map to generate a lossy feature map, perform lossless compression on the lossy feature map to generate a resultant feature map, and store the resultant feature map in the memory.

Abstract: A bloom filter and an implementation method thereof are provided. The implementation method has a ternary rule encoded as a rule binary codeword according to a predetermined encoding rule; has a packet encoded as at least one packet binary codeword according to the predetermined encoding rule; and comparing the rule binary codeword and the at least one packet binary codeword to decide a following processing of the packet. The predetermined encoding rule includes: tagging 0 or 1 into most significant bit (MSB) of the output binary codeword based on mask length of the input codeword; placing the prefix of the input codeword right after MSB of the output binary codeword; and tagging a string to last bits of the output binary codeword based on the mask length, and the bit number of the string equals to the mask length.

Abstract: A hierarchical beamforming method is provided, which comprises the following steps. Firstly, the vectors of the transmitter and the receiver are hierarchically indexed by using the codebooks with different coding methodology to generate hierarchical coding structures. Then, the corresponding vector fields between the transmitter and the receiver are switched according to the hierarchical coding structures. Thereafter, at least two aligning vector fields with greatest gain are detected. Finally, the corresponding aligning index pairs are used to build the channels.

Abstract: A hierarchical beamforming method is provided, which comprises the following steps. Firstly, the vectors of the transmitter and the receiver are hierarchically indexed by using the codebooks with different coding methodology to generate hierarchical coding structures. Then, the corresponding vector fields between the transmitter and the receiver are switched according to the hierarchical coding structures. Thereafter, at least two aligning vector fields with greatest gain are detected. Finally, the corresponding aligning index pairs are used to build the channels.

Abstract: Exemplary embodiments of the present disclosure propose multiple codebook searching methods for a receiver in a communication system, wherein the communication system uses a codebook having L pre-coding matrices. First, L matrices are calculated according to the L pre-coding matrices, and the L matrices are inversed mean square error matrices of the L pre-encoding matrices. Then, diagonal cofactors or determinants of at least S matrices of the L matrices are calculated, and a pre-coding matrix used by the receiver is found according to the maximum cofactors of the diagonal of the at least S matrices or ratios between the maximum cofactors of the diagonals and the determinants of the at least S matrices.

Abstract: A method and apparatus for turbo code decoding are provided to reduce memory consumption during calculation of state metrics. In an embodiment of a turbo code decoder, a natural recursion unit comprises a plurality of add-compare-select (ACS) units performing natural recursion operations to generate a state metric. The original state metric is then converted to a differential metric before being stored into a memory device. The differential metric contains less data than the state metric so that memory consumption is reduced. To restore the original state metric from the differential metric, a plurality of revival units operating in parallel is provided. Thereby, the state metric is reacquired from the differential metric, and a Log Likelihood Recursion (LLR) operation is accordingly performed by an LLR unit.

Abstract: Exemplary embodiments of the present disclosure propose multiple codebook searching methods for a receiver in a communication system, wherein the communication system uses a codebook having L pre-coding matrices. First, L matrices are calculated according to the L pre-coding matrices, and the L matrices are inversed mean square error matrices of the L pre-encoding matrices. Then, diagonal cofactors or determinants of at least S matrices of the L matrices are calculated, and a pre-coding matrix used by the receiver is found according to the maximum cofactors of the diagonal of the at least S matrices or ratios between the maximum cofactors of the diagonals and the determinants of the at least S matrices.

Abstract: An address generator of a communication data interleaver and a communication data decoding circuit are provided. The address generator includes a first operation unit and a second operation unit. The first operation unit receives a first parameter and a first operation result. The first operation unit performs a recursive operation according to the first parameter and the first operation result and outputs the first operation result. The second operation unit receives the first operation result, a second operation result, and a second parameter. According to a transmission mode signal, whether the second operation unit generates a second operation result is determined by performing a recursive operation according to the first operation result, the second parameter, and the second operation result, or by calculating the first operation result and the second parameter.

Abstract: A method and apparatus for turbo code decoding are provided to reduce memory consumption during calculation of state metrics. In an embodiment of a turbo code decoder, a natural recursion unit comprises a plurality of add-compare-select (ACS) units performing natural recursion operations to generate a state metric. The original state metric is then converted to a differential metric before being stored into a memory device. The differential metric contains less data than the state metric so that memory consumption is reduced. To restore the original state metric from the differential metric, a plurality of revival units operating in parallel is provided. Thereby, the state metric is reacquired from the differential metric, and a Log Likelihood Recursion (LLR) operation is accordingly performed by an LLR unit.

Abstract: The Viterbi decoder is an essential module in a communication system, in which the power and the decoding latency are restricted. In the present invention, a power efficient low latency survivor memory architecture and an operating method for the Viterbi decoder are disclosed by providing a plurality of trace-forward units, a plurality of first signal selecting units, a plurality of second signal selecting units and a third signal selecting unit to reduce the power consumption by decreasing the exchange times of contents in the trace-forward units. Thus, the present invention is suitable for use in mobile communication devices which require low power consumption.