This paper presents a novel model-driven deep learning (DL) architecture, called TurboNet, for turbo decoding that integrates DL into the traditional max-log-maximum a posteriori (map) algorithm. The TurboNet inherits...
详细信息
This paper presents a novel model-driven deep learning (DL) architecture, called TurboNet, for turbo decoding that integrates DL into the traditional max-log-maximum a posteriori (map) algorithm. The TurboNet inherits the superiority of the max-log-map algorithm and DL tools and thus presents excellent error-correction capability with low training cost. To design the TurboNet, the original iterative structure is unfolded as deep neural network (DNN) decoding units, where trainable weights are introduced to the max-log-map algorithm and optimized through supervised learning. To efficiently train the TurboNet, a loss function is carefully designed to prevent tricky gradient vanishing issue. To further reduce the computational complexity and training cost of the TurboNet, we can prune it into TurboNet+. Compared with the existing black-box DL approaches, the TurboNet+ has considerable advantage in computational complexity and is conducive to significantly reducing the decoding overhead. Furthermore, we also present a simple training strategy to address the overfitting issue, which enable efficient training of the proposed TurboNet+. Simulation results demonstrate TurboNet+'s superiority in error-correction ability, signal-to-noise ratio generalization, and computational overhead. In addition, an experimental system is established for an over-the-air (OTA) test with the help of a 5G rapid prototyping system and demonstrates TurboNet's strong learning ability and great robustness to various scenarios.
作者：
Park, GHYoon, SHHong, DKang, CEYonsei Univ
Dept Elect & Comp Engn Informat & Telecommun Lab Seodaemoon Gu Seoul 120749 South Korea LG Elect Inc
LG R&D Complex 533 Mobile Comm Res Lab Dept Mobile Comm TechDongan Gu Kyeongku Do 431749 South Korea
Several implementation methods for a map decoder are proposed in this paper, Using a novel pipeline structured time-shared process, the authors are able to efficiently overcome the restrictions imposed by the recursio...
详细信息
Several implementation methods for a map decoder are proposed in this paper, Using a novel pipeline structured time-shared process, the authors are able to efficiently overcome the restrictions imposed by the recursion process on state metrics. and the complexity of the map decoder can be reduced to a level on the order of a SOVA (Soft Output Viterbi algorithm) decoder. In addition. the authors propose an efficient controller structure that can be used for variable frame-size systems such as cdma-2000. The map decoder using a block-wise algorithm designed here was implemented in only one 20,000 gate circuit. It was validated by VHDL, which was compared with the results of the initial simulation (C programs). The decoder demonstrated a 300 kbps decoding processing ability with 8 iterations on a FPGA circuit, with a deviation only about 0.1-0.2 dB greater than that for an ideal map decoder, even when all hardware environments are considered.
In the paper, a new implementation of a 3GPP LTE standards compliant turbo decoder based on GPGPU is proposed. It uses the newest GPU-Tesla K20c, which is based on the Kepler GK110 architecture. The new architecture h...
详细信息
ISBN:
(纸本)9781479944163
In the paper, a new implementation of a 3GPP LTE standards compliant turbo decoder based on GPGPU is proposed. It uses the newest GPU-Tesla K20c, which is based on the Kepler GK110 architecture. The new architecture has more powerful parallel computing capability and we use it to fully exploit the parallelism in the turbo decoding algorithm in novel ways. Meanwhile, we use various memory hierarchies to meet various kinds of data demands on speed and capacity. Simulation shows that our implementation is practical and it gets 76% improvement on throughput over the latest GPU implementation. The result demonstrates that the newest Kepler architecture is suitable for turbo decoding and it can be a promising reconfigurable platform for the communication system.
In the field of mobile communications, the energy issue of a turbo decoder becomes an equivalent constraint as throughput and performance. This paper describes a technique to reduce the internal bitwidth of the state ...
详细信息
ISBN:
(纸本)9781424412211
In the field of mobile communications, the energy issue of a turbo decoder becomes an equivalent constraint as throughput and performance. This paper describes a technique to reduce the internal bitwidth of the state metrics, and hence, to decrease the entire energy dissipation of a turbo decoder. This approach is based on the saturation of the state metrics. Two cases are investigated: saturation outside the ACS recursion loop and saturation inside the ACS recursion loop. The targeted system is the Universal Mobile Telecommunications System (UTMTS) with an 8-state turbo decoder using the max-log-map algorithm. When received symbols and extrinsic informations are respectively 4-bit and 6-bit quantized, the internal bitwidth of the state metrics can be reduced from 7 bits downto 4 bits. This reduction is paid by a loss of 0.1 dB at a Bit Error Rate (BER) of 10(-6). In addition, when 4 SISO decoders perform in parallel, the proposed optimization yields to a reduction of memory area by 10% and leads to an energy reduction of 24% for a 70 mn technology.
This paper proposes a low-complexity design for BCJR or maximum a posteriori (map) decoders. max-log-map algorithm, which offers a good compromise between performance and complexity, is selected for implementation. Th...
详细信息
ISBN:
(纸本)0780391527
This paper proposes a low-complexity design for BCJR or maximum a posteriori (map) decoders. max-log-map algorithm, which offers a good compromise between performance and complexity, is selected for implementation. The decoder architecture is parallel and pipeline. New quantization and new branch metric calculation unit are proposed in this paper. Overall decoder design is flexible to the transmission block lengths, which makes it appropriate for variable length transmission systems. Performance results are tested on turbo decoders and show that the proposed BUR decoder is quite suitable for turbo decoders. The chip is implemented in a 0.18-mu m six-layer metal CMOS technology and has the core size of 1mm(2) (including two memories). The throughput of the turbo decoder is 6.5Mb/s after 10 iterations.
Probabilistic constellation shaping (PCS) improves the spectral efficiency by adjusting the probability distribution of the transmitted symbols to match the channel characteristics. The conventional log-maximum A Post...
详细信息
ISBN:
(纸本)9798350304060;9798350304053
Probabilistic constellation shaping (PCS) improves the spectral efficiency by adjusting the probability distribution of the transmitted symbols to match the channel characteristics. The conventional log-maximum A Posteriori (log-map) algorithm can work for demodulation of PCS system by taking into account the probability of each constellation point. However, the conventional max-log-map algorithm cannot be applied directly. In this paper, a simplified soft-demapping scheme is proposed based on the max-log-map algorithm for the PCS system. The proposed scheme has a lower complexity than the logmapalgorithm with acceptable performance loss. In addition, the distribution mismatch between transmitter and receiver is evaluated. The results show that four distributions are adequate for PCS-64 and PCS-256 to cover a wide range of signal-to-noise ratios.
We consider high-rate systematic recursive convolutional encoders to be adopted as constituent encoders in turbo schemes. It has been shown by Douillard and Berrou that the construction of high-rate turbo codes by mea...
详细信息
ISBN:
(纸本)9781457705953
We consider high-rate systematic recursive convolutional encoders to be adopted as constituent encoders in turbo schemes. It has been shown by Douillard and Berrou that the construction of high-rate turbo codes by means of high-rate constituent encoders offers several advantages over the typical construction based on the puncturing of rate-1/2 constituent encoders. To reduce the decoding complexity associated with high-rate codes, we adopt the "minimal" trellis representation of convolutional codes introduced by McEliece and Lin. While in the literature this trellis has been obtained for nonrecursive nonsystematic generator matrices, we herein introduce the construction of the "minimal" trellis for a systematic recursive convolutional encoding matrix. We also derive expressions for the arithmetic decoding complexity when the max-log-map algorithm is applied over the conventional and the "minimal" trellises. Examples are provided, which show that significant savings in decoding complexity are obtained, while keeping the same error performance of conventional schemes, when the minimal trellis is used. Finally, a code search is conducted and examples are provided which indicate that a refinement in terms of decoding complexity-error performance trade-off is obtained.
This paper focuses on implementing an efficient and simplified turbo decoder with FPGA. The max-log-map algorithm is employed and a simplified and improved max-log-map algorithm is present. By carefully manipulating h...
详细信息
ISBN:
(纸本)9781424421077
This paper focuses on implementing an efficient and simplified turbo decoder with FPGA. The max-log-map algorithm is employed and a simplified and improved max-log-map algorithm is present. By carefully manipulating hardware, we implement the whole turbo decoder with a single-RSC structure. Comparing with the conventional decoder, our turbo decoder reduces about 60% hardware cost and operates up to 2Mbit/s. Finally, we investigate the performance of our FPGA-based turbo decoder in a FH-SS system with partial band jamming. The results drawn from this code are compared with a RS code. Results show that the turbo code significantly outperforms the RS code.
In this paper, the performance of the Least Squares (LS) estimator along with turbo codes in OFDM is evaluated for the SUI channel models in the presence of AWGN. Turbo codes are used in this paper because of its BER ...
详细信息
ISBN:
(纸本)9781479925810
In this paper, the performance of the Least Squares (LS) estimator along with turbo codes in OFDM is evaluated for the SUI channel models in the presence of AWGN. Turbo codes are used in this paper because of its BER performance close to Shannon limit. The iterative turbo decoder exchanges soft information between the component decoders. maximum A Posteriori (map) and its simplified version max-log-map algorithms are used for the component decoders. The performance and complexity comparison is also made between the decoding algorithms. Moreover the effect of memory order on the performance of the decoding algorithms is also studied.
The max-log-map(MLM) algorithm is simple for Turbo-codes' decoding, based on the algorithm a structure of parallel array for Turbo-codes decoding is suggested,which is easy to implement by *** this structure,flow ...
详细信息
ISBN:
(纸本)0780363949
The max-log-map(MLM) algorithm is simple for Turbo-codes' decoding, based on the algorithm a structure of parallel array for Turbo-codes decoding is suggested,which is easy to implement by *** this structure,flow direction of dates and computation process of the algorithm are obtained,the relation of nodes and date frames are analyzed,and the simple time sequencing of dates operation is *** structure is proved by simulation.
暂无评论