Single-cell RNA-seq enabled in-depth study on tissue micro-environment and immune-profiling, where a crucial step is to annotate cell identity. Immune cells play key roles in many diseases, whereas their activities are hard to track due to their diverse and highly variable nature. Existing cell-type identifiers had limited performance for this purpose. We present HiCAT, a hierarchical, marker-based cell-type identifier utilising gene set analysis for statistical scoring for given markers. It features successive identification of major-type, minor-type and subsets utilising subset markers structured in a three-level taxonomy tree. Comparison with manual annotation and pairwise match test showed HiCAT outperforms others in major- and minor-type identification. For subsets, we qualitatively evaluated the marker expression profile demonstrating that HiCAT provide the clearest immune-cell landscape. HiCAT was also used for immune-cell profiling in ulcerative colitis and discovered distinct features of the disease in macrophage and T-cell subsets that could not be identified previously.

We consider the joint MIMO detection algorithm based on ring-type Pair-wise Markov random fields (MRF) in [1] for Gaussian input to obtain the Gaussian BP that is run over the ring-type MRF and proves its convergence in mean to the linear MMSE estimates. The result lies on the same line of those in [2] and [3], but with differences in its graphical model and the message passing rule.

Low complexity, iterative MIMO detection algorithms are derived based on pair-wise Markov random fields (MRF). We consider two types, namely, the fully-connected and the ring type MRF and, for the edge potentials, we use the bivariate Gaussian function obtained by marginalizing the posterior joint probability density under Gaussian-input assumption. Since the corresponding factor graphs has only 2 edges per factor node, the computations are much easier than that of ML which is similar to the belief propagation algorithm run over the fully connected factor graph. Compared to the complexity of ML, O(2mM), the proposed scheme for the fully-connected and the ring-type MRF is shown to have only O(M ·(M-1) 2m-1) and O(M·2m), respectively, for M being the input dimension and m the number of bits per data symbol. The performances are evaluated, via simulation, in terms of bit error rate with DVB-S2 LDPC coding.

In this paper, we propose a hybrid beam-forming and beam-switching technique for OFDM based wireless personal area networks (WPAN). In order to compromise between performance and computational cost and overheads for the channel side information feedback, the proposed system employs hybrid beam-forming and beam-switching, where the block beam-switching using a predefined beam codebook is used at the transmitter while the optimum per-subcarrier beam-forming is used at the receiver. To verify the performance of the proposed scheme, the effective SNR gain over the single antenna transmission system is investigated along with the spectral efficiency bound for some of the channels developed by the IEEE 802.15.3c task group. For comparison, the performance of the optimum per-subcarrier beam-forming and the block beam-switching are also examined.

A MIMO decoding algorithm utilizing poly- diagonalization and tail-biting trellis is proposed. Linear MIMO decoding, such as zero-forcing equalization, is a channel diagonalization technique, where all interferences from other data streams are suppressed for separate decoding. It is well known, however, that such decoders suffer from the noise enhancement. In the proposed scheme, interferences are partially allowed, in order to alleviate the noise enhancement and, then, post trellis decoding is used for the joint detection utilizing the poly-diagonal structure of the effective channel. Under the proposed framework, the zero-forcing equalizer can be regarded as special cases of poly-diagonalization, i.e., of the first order. And the proposed scheme can provide a tradeoff between the complexity and performance by choosing an appropriate order of poly-diagonalization. According to the simulation results, considerable gain can be obtained even with the second order, i.e., bi-diagonalization, for which the decoding complexity is far less than that of the maximum likelihood decoding.

A MIMO detection algorithm utilizing poly- diagonalization and tail-biting trellis is proposed. Linear MIMO detection, such as zero-forcing or minimum mean squared error (MMSE) equalization, is basically a channel diagonalization technique, where interferences from other data streams are decoupled for the separate decoding. It is well known, however, that such decoders suffer from noise enhancement, which causes considerable performance degradation. In this paper, we propose poly-diagonalization of the channel matrix in zero-forcing and MMSE senses. The idea behind poly-diagonalization is to allow interferences partially in order to alleviate the noise enhancement and, then, to use post trellis decoding for the joint detection utilizing the poly-diagonal structure of the effective channel. Under the proposed framework, the zero-forcing and the MMSE equalizer can be regarded as special cases of poly-diagonalization, i.e., of the first order. The proposed scheme can provide a tradeoff between the complexity and performance by choosing an appropriate order of poly-diagonalization. According to the simulation results, considerable gain can be obtained even with the second order, i.e., bi-diagonalization, for which the decoding complexity is far less than that of the maximum likelihood decoding.

The dynamic resource allocation and throughput behavior of a sub-channelized multiuser orthogonal frequency division multiplexing (OFDM) system are considered. In a sub- channelized OFDM, a set of sub-carriers are reserved for use as the unit of channel quality information (CQI) feedback, user- multiplexing and power allocation. To design and analyze the sub-channelized OFDM system, we first model the empirical SNR distribution over a sub-channel as Ricean and develop a two-step resource allocation algorithm that utilizes the estimated Ricean parameters as the statistical CQI. For two generic subchannel structures, namely the contiguous and the distributed, we explore the connection between the sub-channel structure and the network throughput in a sub-channelized OFDM.

System level performance of a broadcast/unicast superposition coding system is considered. Cellular network for unicast service only is considered as interference limited system, where increasing transmission power does not help improving network throughput. It is the case especially when frequency reuse factor is close to 1. In such case, the amount of power that does not contribute to improving throughput can be considered as "unused." This situation motivates to use the unused power for broadcast service, which can be efficiently provided in OFDM based single frequency network (SFN) as of the digital video/audio broadcast system. In this paper, we investigate the performance of such an overlay system where single frequency broadcast network is superimposed over unicast cellular service, namely, BC/UC superposition coded system (supercast). Alternative service multiplexing using FDM/TDM will also be investigated for comparison.

In this paper, we consider downlink throughput performances of multiuser orthogonal frequency division multiplexing (multiuser OFDM) with reduced channel information feedback schemes. Specifically, two types of reduced feedback schemes, namely, 1-bit per sub-carrier and selective feedback scheme are considered and compared with each other in terms of average network throughput. For the latter, since the exact analysis is complicated, we resort to an approximate analysis. Simulations results will also be provided to verify the approximate analysis. Since the strict throughput comparison for given number of feedback bits per user is quite difficult, rather we compare their general behaviors in various system configurations with different system parameters, which can give us an insight into practical system design with those reduced feedback schemes.

In this paper, a receiver algorithm for an 8times8 MIMO system is considered for Giga-bps data transmission, in which the decoding complexity and latency are crucial factors for implementation of MIMO system with large number of transmit and receive antennae. To reduce computational complexity and the decoding latency, a 3-stage receiver algorithm is proposed. The MIMO receiver consists of multi-dimensional detection, partial interference cancellation and the weighted zero-forcing equalization, where the transmit antennae are divided into two group. The first group is encoded by a lower rate code and multi-dimensional search algorithm is used for demodulation and decoding, while the second group is encoded by higher rate code and weighted zero-forcing equalization is used. Once the first group is decoded successfully, they are cancelled out from the received vector for the decoding of the second group. The complexity and the error rate performance are investigated and compared with a well-known sphere decoding algorithm.

In this paper, we propose a design methodology of low-density parity check (LDPC) code for very high speed transmission. To support various code rates with reasonable hardware and, possibly, the type-II hybrid automatic repeat request (HARQ), a row-splitting/combining method is employed in conjunction with shortening. In row-splitting approach, a parity check matrix of a high rate mother LDPC code, saying rate 5/6, is designed first and then successively row-split the mother matrix to obtain lower rate codes. Shortening provides us another degree of freedom for rate and frame length flexibility for concatenation with MIMO system. The code performance is investigated in terms of frame error rate and compared with those of the LDPC code proposed in IEEE 802.16e standard.

A novel "permuted sequences combination" method is proposed in this paper, which uses the conventional frame structure wherein two sync words, each comprising M complex symbols, are appended at the frame start. The goal of this new method is to reduce the large variance of the timing estimation error inherent in the conventional correlation method. Essentially, the first sync word, stilde1, is a predetermined constant amplitude zero autocorrelation (CAZAC) sequence. The second sync word, stilde2, which is a certain permutation of the first so that the combination of the two received sync signal vectors perform sliding window processing with peak occurs at the correct frame start. We show that the symmetric correlation method previously presented in (Park et al., 2003) for orthogonal frequency division multiplexing (OFDM) systems can be regarded a special case of the proposed permuted sequences combination method. As a special case of permuted sequences combination method, symmetric correlation method is robust to the frequency offset. The proposed methods have good simulation results in both AWGN channel and multipath environments. Simulation results support the analysis.

Automatic repeat request (ARQ) is one of the effective techniques against the transmission error at the medium access control (MAC) layer. An erroneous MAC protocol data unit (MPDU) can be typically retransmitted within a given limit. In order to improve the IP-level performance, which directly affects the user-perceived quality-of-service (QoS), we propose a different ARQ strategy, called 'MSDU-based' scheme, with simple and minor modification from the existing ARQ concept. We evaluate the analytical IP-level performance of both the conventional and proposed scheme.

Detection performance is analyzed for pilot-aided base station acquisition in OFDM based cellular networks. With non-coherent decision statistics, we find the slope of the pilot pattern, which is evenly spaced in the frequency domain and hops from OFDM symbol to OFDM symbol. A pilot symbol location is defined as a set of Latin square sequences with different (evenly spaced) offset values. We examine the probability of correct detection with various parameter sets, such as pilot-to-signal power ratio, frequency offset and observation window size.

System level performance of OFDM/ OFDMA is evaluated for various system parameters, such as frequency reuse factor, scheduling algorithm and sub-channel structures. Specifically, the generalized two step sub-channel/power allocation algorithm, with either aggregated or distributed sub-channel structure, is first extended to provide proportional fairness. Then, by simulation, the sector throughput is evaluated in practical OFDMA based cellular system environment with frequency reuse of 1 and 3, respectively.

In this two part paper, we consider a packet data communication over code-division multiple-access (CDMA) system. As a preliminary, this part of the paper highlights the practically achievable spectral efficiency of a CDMA system with turbo codes and multiuser detection, assuming the use of hybrid type-I automatic-repeat-request. We also investigate the effect of code rates on the performance of systems with matched filter and linear minimum mean square error (MMSE) receivers, respectively. For a given ensemble distance spectra of punctured turbo codes, we use an improved union bound to evaluate the error probability of maximum likelihood turbo decoder with matched filter receiver and with linear MMSE receiver front-end, and then compute the corresponding throughput and spectral efficiency as a function of system load. The results show that with 1/3 rate turbo code and matched filter receiver a spectral efficiency of around 1.1 b/s/Hz can be obtained, while, by employing a linear MMSE receiver, the spectral efficiency can be improved further up to 1.6 b/s/Hz. The results also show that the spectral efficiency with a linear MMSE receiver saturates at a code rate of about 4/5, i.e., does not improve further with lower rate coding, while, for matched filter receivers, better performance can be obtained with lower code rate. Theoretically, as the system load goes to infinity, the spectral efficiency of linear MMSE receiver converges to that of matched filter receiver. In a practical situation, however, there exists a noticeable performance gap between the two receivers unless a very low rate code is used.

In this part of the paper, we investigate the throughput bound of CDMA unslotted ALOHA system with a hybrid type II automatic-repeat-request (ARQ) implemented using rate compatible punctured turbo codes. The system under consideration can be regarded as an enhanced version of the classical CDMA packet radio network. When applying a hybrid type-II ARQ, the probability of packet success and packet length is generally different from attempt to attempt. Since the conventional analytical model, customarily employed for the ALOHA system with pure or hybrid type-I ARQ, cannot be applied for this case, we introduce an expanded analytical model, which can be regarded as a network of queues, and apply Jackson and Burke's theorems to simplify the analysis. When using code combining, soft-decision decoding is usually considered, in which case, it is very difficult to find the actual probability of packet success, especially when the number of terminals changes during a packet time duration. Hence, instead of finding the actual probability of packet success, we evaluate a lower and an upper bound. Together with the expanded model and the packet success probability bound, we obtain the corresponding throughput bounds by iteratively searching the steady-state arrival rate, /spl lambda/, and the average departure rate, /spl mu/, for given composite offered traffic, G=/spl lambda///spl mu/. The results show that significant improvement in throughput can be achieved by exploiting high coding gain of turbo codes and channel utilization efficiency of incremental redundancy retransmission protocol.

We derive a throughput bound for CDMA unslotted ALOHA with memory ARQ in general. Since the conventional 'single queue with feedback' model, customarily employed for an ALOHA system with pure or hybrid type-I ARQ, cannot be applied to this case, we introduce an expanded model, which consists of a network of queues, and apply the Jackson theorem to simplify the analysis. When using code combining, it is very difficult to find the actual probability of packet success. Hence, instead of finding actual values, we evaluate a lower bound on it. Together with the expanded model and the packet success probability bound, we obtain the throughput bound for given composite offered traffic by using an iterative search procedure.

We propose a construction scheme of composite QAM sequence by weighted sum of the QPSK sequences that have relatively low peak-to-mean envelope power ratio (PMEPR). We derive a general expression of the upper bound on PMEPR for such composite QAM sequences. For the proposed QAM sequence construction, which we call 'linear-sum mapping', we investigate the information transfer capability. Although it is inferior to Gray mapping, by utilizing the decisions of the previously decoded codeword and employing successive decoding, we can obtain almost the same performance as that of Gray mapping.

We analyze the system capacity of turbo coded CDMA packet system with hybrid FEC/ARQ scheme. In particular, assuming the use of punctured turbo codes, we investigate the system capacity in terms of the number of users that can be supported by the given rate of punctured turbo codes in both the AWGN and multipath fading channel. For a multipath fading channel, we derive a simple closed form expression for the PDF of signal to noise plus interference ratio (SNIR) at the output of matched filter receiver. Then, using an improved union bound, we evaluate the codeword error probabilities and the corresponding throughput performance for various rates of punctured turbo codes. With these results, we obtain the CDMA system capacity and compare its performance with theoretical capacity bounds obtained with idealized packet communication systems in which packet overheads, transmission and decoding delay can be ignored.

The performance of a linear decorrelating detector (LDD) and a minimum mean square error (MMSE) detector is analyzed for random spreading waveforms. The performance of the LDD and MMSE detectors is expressed in terms of the so-called near-far resistance, defined by a reciprocal of a diagonal component of inverse matrix. For random code division multiple access, which employs random spreading waveforms, the near-far resistance can be regarded as a random variable. Many papers have dealt with the analysis of multiuser detectors for random spreading sequences. In most cases, however, these analyses derived only the expectations or bounds for the near-far resistance. In this paper, we directly derive the approximate probability density function (PDF) of the near-far resistance and corresponding bit error rate expression for random spreading sequences. It is based on Gaussian approximation of the cross correlation between any two randomly generated spreading codes. The resulting PDF turned out to be a reversed-and-scaled version of chi-square distribution. The approximate expressions, both the PDF and the corresponding bit error rate expression, were verified via Monte Carlo simulations. The results showed that the approximation is quite close to the simulation results when the number of users is less than half the processing gain.