Search

Small-scale magNetosphere and Ionosphere Plasma Experiment (SNIPE) mission is aimed to observe a small-scale structure of the physical phenomena in a near-Earth environment. SNIPE mission comprises four 6U size nanosats that perform formation flying to meet the scientific objectives of the mission. In this study, we designed a formation flying system and validated it using numerical simulations to collect temporal and spatial differences of the physical phenomena for the SNIPE mission. The requirements for spacecraft formation flying are to pass the same point at different times and be located at different longitude points on the same latitude. Two types of formations which are an along-track formation for temporal observations and a cross-track formation for spatial observations are devised. The size of the formation decreases during the along-track formation phase and increases during the cross-track formation phase. Four types of orbit control were introduced to implement these formations and adjust the relative distances between the nanosats. Changing the shape and size of the formation exploits the effects of perturbations through orbit controls. The simulations were considered with several constraints to replicate the intended environments. The simulations also reflect the time differences between the orbit determination epoch and orbit control epoch, conditions in attitude control of nanosats, limitations of orbit control operation time to secure power stability, and errors in the thrust module. The numerical simulations demonstrated that each nanosat satisfied the mission requirements in a Δ V budget of 50 m/s. These results verified that the designed formation flying system meets the scientific objectives of SNIPE mission by changing the shape and size of the formation.

The purpose of this study was to investigate the effects of three back-support exoskeletons (FLx ErgoSkeleton, V22 ErgoSkeleton, Laevo V2.5) and patient transfer methods (Squat pivot, stand pivot, scoot) on the musculoskeletal loading and self-reported usability measures during patient transfers between a bed and a wheelchair. In a repeated-measures laboratory study, 20 experienced caregivers (17 females and 3 males) performed a series of 24 bed-to-wheelchair transfer tasks (2 directions × 4 exoskeleton conditions × 3 patient transfer methods). The trunk flexion and lateral flexion angles, bilateral hand pull forces, and muscle activities of the erector spinae were significantly different by exoskeleton conditions and patient transfer methods (p’s < 0.01). The usability measures were significantly affected by exoskeleton designs (p’s < 0.01). There were significant two-way interaction effects on the trunk flexion and lateral flexion angles and muscle activities of the erector spinae (p’s < 0.01). For the squat pivot method, three back-support exoskeletons showed the largest reduction of erector spinae muscle activities (47.4 to 83.5% reference voluntary contractions) compared to no exoskeleton. This indicated the effects of exoskeleton conditions on the trunk postures and erector spinae muscle activities depended on the patient transfer method. More research could be needed to improve the trunk postures and usability of back-support exoskeletons suitable for patient handling.

We describe a method for the in-orbit calibration of body-mounted magnetometers based on the CHAOS-7 geomagnetic field model. The code is designed to find the true calibration parameters autonomously by using only the onboard magnetometer data and the corresponding CHAOS outputs. As the model output and satellite data have different coordinate systems, they are first transformed to a Star Tracker Coordinate (STC). Then, non-linear optimization processes are run to minimize the differences between the CHAOS-7 model and satellite data in the STC. The process finally searches out a suite of calibration parameters that can maximize the model-data agreement. These parameters include the instrument gain, offset, axis orthogonality, and Euler rotation matrices between the magnetometer frame and the STC. To validate the performance of the Python code, we first produce pseudo satellite data by convoluting CHAOS-7 model outputs with a prescribed set of the ‘true’ calibration parameters. Then, we let the code autonomously undistort the pseudo satellite data through optimization processes, which ultimately track down the initially prescribed calibration parameters. The reconstructed parameters are in good agreement with the prescribed (true) ones, which demonstrates that the code can be used for actual instrument data calibration. This study is performed using Python 3.8.5, NumPy 1.19.2, SciPy 1.6, AstroPy 4.2, SpacePy 0.2.1, and ChaosmagPy 0.5 including the CHAOS-7.6 geomagnetic field model. This code will be utilized for processing NextSat-1 and Small scale magNetospheric and Ionospheric Plasma Experiment (SNIPE) data in the future.

Recently, flexible pressure sensors have gained substantial research interest in bioelectronics because they can monitor the conditions of various organs, enable early diagnosis of diseases, and provide precise medical treatment by applying them to various parts of the body. In particular, inorganic materials, metal and carbon-based materials are broadly used in novel structured pressure sensors from wearable devices to implantable devices. With the excellent electronic properties, distinctive morphologies, and remarkable mechanical and chemical stability of these materials, it is expected that these flexible pressure sensors can be the basis for new methods for human healthcare. This article covers an extensive review of the inorganic, metal and carbon-based flexible pressure sensor design strategies and sensing mechanisms studied in recent years for diverse applications such as tactile sensors, arterial pulse sensors, intracranial pressure sensors, intraocular pressure sensors, and bladder pressure sensors. Each section provides an overview by introducing the recent progress in flexible pressure sensors.

To meet the demand for storing large amounts of data, there has been a stronger focus on increasing the recording density of a hard disk drive (HDD). While conventional HDD suffers from thermal effect and superparamagnetic limit, bit-patterned media recording (BPMR) has the potential to resolve these problems and is capable of increasing areal density (AD) beyond 1 terabit per square inch. To increase the AD of BPMR, the distance between islands in both the down- and cross-track directions must be reduced. Consequently, the decreased bit period and track pitch induce more intersymbol interference (ISI) and intertrack interference (ITI), which degrade the bit error rate (BER) performance. In this study, extrinsic information obtained from a multilayer perceptron (MLP) equalizer is used as a priori information in a partial response maximum-likelihood (PRML) detector to improve the detection performance of BPMR. The proposed detection scheme shows better performance compared to both the MLP equalizer and conventional PRML alone. In addition, it was found that the proposed detector provides improved performance when track misregistration (TMR) occurs.

This study investigated the effect of different hologram distances on the shoulder postures, muscle activities, and the task performance (speed) during Augmented Reality (AR) interactions. With a repeated- measures design, 20 participants (10 males) performed the 3-D cube task involving gaze, pinching, and dragging gestures by different hologram distances: near (30 cm), middle (60 cm), and far (90 cm) relative to the participant. Meanwhile, muscle activities (upper trapezius, anterior and middle deltoid) and the right shoulder postures were measured. The results showed that hologram distances significantly affected middle deltoid muscle activities and shoulder flexion and abduction angles, and task performance during 3-D cube task ( p’s < 0.001). The far hologram distance resulted in higher middle deltoid muscle activity and shoulder angles compared to the near and middle distances. Task performance was higher when the hologram was displayed in the middle compared to the near and far distances ( p’s < 0.001). These results indicate that a hologram distance may be an important design factor affecting users’ shoulder stress and task performance during AR interactions.

Sparse arrays have been studied mainly to reduce the large numbers of elements in 2-D arrays. However, they can also provide an effective means of miniaturizing ultrasound 1-D array systems for point-of-care applications. Although a variety of sparse array design strategies have been proposed, designing an optimum sparse array to simultaneously satisfy the system specification requirements and performance criteria remains a challenge. This article presents an analytical approach for the design of an optimum pair of periodic sparse arrays (PSAs), one for transmission and the other for reception. The approach is based on three newly derived theorems that describe the most important properties of the two PSAs forming the sparse array pair and their relationship pertaining to the overall beam pattern. The proposed approach can be used to design 1-D sparse array pairs with arbitrary sparseness factors while meeting given performance criteria. The computer simulation verified that the spatial resolution of a 64-element phased array can be obtained with a PSA pair consisting of transmit and receive sparse arrays, of which the number of elements is reduced to 32 and 22, respectively.

Holographic data storage (HDS) can access data in a short amount of time and can also quickly transmit data because HDS records and reads data in units of pages. Thus HDS is regarded as a promising candidate for future optical recording storage systems. Conventional storage systems such as CD, DVD, and HDD have one-dimensional (1D) inter-symbol interference (ISI). However, HDS has a 2D ISI that degrades performance. In this study, to avoid and overcome the 2D-ISI problem, the authors propose a 14/16 modulation code for 4-ary HDS systems which represents 2 bits on each pixel. The proposed 14/16 modulation code allows a bit of unwanted ISI cases in order to raise the code rate, but it has a higher code rate but similar performance compared to a 3/4 modulation code that has no unwanted ISI cases.

Acoustic-sensor-based soft error resilience is particularly promising, since it can verify the absence of soft errors and eliminate silent data corruptions at a low hardware cost. However, the state-of-the-art work incurs a significant performance overhead for in-order cores due to frequent structural/data hazards during the verification. To address the problem, this paper presents Turnpike, a compiler/architecture co-design scheme that can achieve lightweight yet guaranteed soft error resilience for in-order cores. The key idea is that many of the data computed in the core can bypass the soft error verification without compromising the resilience. Along with simple microarchitectural support for realizing the idea, Turnpike leverages compiler optimizations to further reduce the performance overhead. Experimental results with 36 benchmarks demonstrate that Turnpike only incurs a 0-14% run-time overhead on average while the state-of-the-art incurs a 29-84% overhead when the worst-case latency of the sensor based error detection is 10-50 cycles.

In a bit-patterned media recording (BPMR) system, reducing the spacing between adjacent bit islands in the down- and cross-track directions increases areal density; however, two-dimensional (2-D) interference comprising intersymbol and intertrack interferences degrades the system performance. To mitigate error patterns that increase the 2-D interference in BPMRs, 2-D modulation coding schemes have been proposed. In this letter, we present a modulation code and its decoding using a multilayer perceptron (MLP) for BPMRs. Modulation decoding using an MLP can achieve significantly better performance than that using the minimum Euclidean distance, which is generally used to decode received modulated data. Similarly, when comparing the system performance at the same user density, modulation decoding using an MLP at 3 Tb/in2 shows better performance than an uncoded system (partial response maximum-likelihood detector) at 2 Tb/in2.

Bit-patterned media recording (BPMR) is a candidate technology that extends the density to up to 4 Tb/in2 to handle large amounts of data. However, in BPMR systems, the major challenge is two-dimensional (2-D) interference, which includes intersymbol interference from the down-track direction and intertrack interference from the across-track direction. To combat 2-D interference, it is natural to consider a 2-D Viterbi detection scheme. Since the complexity of pure 2-D Viterbi detection is quite high, we propose a series of two 1-D Viterbi detectors, which is based on new 2-D partial response (PR) target modeling for the BPMR channel. The use of new 2-D PR target model and the two serial 1-D Viterbi detection scheme ensures that the complexity is greatly reduced compared to 2-D Viterbi detection and the bit error rate performance is improved.

This paper presents designs, multiphysics finite-element method (FEM) and circuit simulations, and experimental validations of a three-phase inverter power module for an electric compressor (e-Compressor) used in 48-V mild hybrid vehicles. The inverter module is integrated with its motor and scroll. The integration and e-Compressor operation situations raise sequential challenges in the module design and analysis. Thus, the e-Compressor module employs a direct bonded copper (DBC) substrate on which MOSFET dies, shunt resistors, and other electronics are soldered. Multiphysics FEM and circuit simulations are conducted to analyze (and minimize) the parasitic and loop inductances, current concentration, and temperature elevation of the design. The modeling and simulation processes are customized to represent the integrated DBC-based e-Compressor module, including the determination of loop inductance and overshoot voltage by considering systemic current conduction and reasonable simplification of cooling fluid effects. The simulation process and module performances are validated by experiments. The designed e-Compressor inverter module is improved to fully satisfy all the required specifications. Moreover, an excellent agreement between the simulation and the experimental results is confirmed, demonstrating the similar reductions of: 1) loop inductances simulated by FEM (~57%); 2) circuit-simulated overshoot voltages (~47%); and 3) experimentally measured overshoot voltages (~53%).

In this study, we propose an approach for the design and satisfy the requirements of the fabrication of a reliable and stable high-frequency downconverter for the millimeter-wave (Ka band) and detail the contents of the approach. We design and fabricate a stable downconverter with a low noise figure, flat gain characteristics, and multi-channel characteristics suitable for millimeter-wave bands. The method uses the chip-and-wire process for the assembly and operation of a bare MMIC device into the RF path. To compensate for the mismatch among the many components used in the module, W/G transition, an image rejection mixer, a switch, and an amplifier suitable for millimeter-wave frequency characteristics are designed and applied to the downconverter. To reject the spurious signals generated from the complex local oscillation signals, the downconverter is designed to not affect the RF path. In the Ka-band downconverter, the gain is measured from 41.89 dB to 42.83 dB at 33–35 GHz with flatness of about 0.94 dB. The measured value of the noise figure at CH1 is 4.936 dB with a maximum value in the 0.75–1.25 GHz intermediate frequency. The third intermodulation measurement result is 61.83 dBc under a −50 dBm input power and above gain, and the switching to select a channel takes about 622 μs.

Since conventional data storage systems are faced with difficult problems such as superparamagnetic limit and thermal effect, bit-patterned media recording (BPMR) which resolves these problems receives attention as a promising next generation magnetic data storage system. The advantages of BPMR are reduced track edge noise and nonlinear bit shift and easily simplified tracking. However, the reduced spacing between islands to increase areal density causes severe intersymbol interference (ISI) and intertrack interference (ITI) which degrade BPMR system performance. To prevent error patterns that bring more ITI, modulation coding schemes was proposed [1] . Also, to improve system performance, neural networks have been applied to detection and decoding in data storage system [2] . In this paper, we propose two serial multi-layer perceptrons (MLPs) which are used for signal detection and modulation decoding, respectively, for BPMR.

This paper proposes a framework called SnuRHAC, which provides an illusion of a single GPU for the multiple GPUs in a cluster. Under SnuRHAC, a CUDA program designed to use a single GPU can utilize multiple GPUs in a cluster without any source code modification. SnuRHAC automatically distributes workload to multiple GPUs in a cluster and manages data across the nodes. To manage data efficiently, SnuRHAC extends CUDA Unified Memory and exploits its page fault mechanism. We also propose two prefetching techniques to fully exploit UM and to maximize performance. Static prefetching allows SnuRHAC to prefetch data by statically analyzing CUDA kernels. Dynamic prefetching complements static prefetching. SnuRHAC enforces an application to run on a single GPU if it is not suitable for multiple GPUs. We evaluate the performance of SnuRHAC using 18 benchmark applications from various sources. The evaluation result shows that while SnuRHAC significantly improves ease-of-programming, it shows scalable performance for the cluster environment depending on the application characteristics.

Widely used Deep Learning (DL) frameworks, such as TensorFlow, PyTorch, and MXNet, heavily rely on the NVIDIA cuDNN for performance. However, using cuDNN does not always give the best performance. One reason is that it is hard to handle every case of versatile DNN models and GPU architectures with a library that has a fixed implementation. Another reason is that cuDNN lacks kernel fusion functionality that gives a lot of chances to improve performance. In this paper, we propose a DL optimization framework for versatile GPU workloads, called DeepCuts. It considers both kernel implementation parameters and GPU architectures. It analyzes the DL workload, groups multiple DL operations into a single GPU kernel, and generates optimized GPU kernels considering kernel implementation parameters and GPU architecture parameters. The evaluation result with various DL workloads for inference and training indicates that DeepCuts outperforms cuDNN/cuBLAS-based implementations and the state-of-the-art DL optimization frameworks, such as TVM, TensorFlow XLA, and TensorRT.

Text-to-speech (TTS) models trained to minimize the spectrogram reconstruction loss can learn speaker embeddings without explicit speaker identity supervision, unlike x-vector speaker identification (SID) systems. Leveraging this way of speaker embedding learning can be useful in unsupervised or semi-supervised scenarios where non, or only some, of the training data have speaker labels. Thus, in this paper, we evaluate speaker embeddings learned by training the spectrogram prediction network under unsupervised and semi-supervised scenarios. We experimented with different data sampling strategies. The best one was sampling two different segments from the same utterance, namely A and B, where the spectrogram of B is predicted given the B phone sequence and the speaker embedding extracted from A. This method improved by 3.4% relative in EER, compared to using the same utterance for both A and B without segmenting. In the unsupervised scenario, the best speaker embedding outperformed i-vectors, the state-of-the-art unsupervised speaker embedding, in speaker verification by 12.9% relative in EER. We observed high correlation between reconstruction loss and speaker embedding quality. In the semi-supervised scenario, having more unlabeled data in training led to a better performance in speaker verification. Adding 5314 unlabeled speakers to 800 labeled speakers improved EER by 10.8 % relative.

Ultra-high density data storage has gained high significance given the increasing amounts of data, many technologies have been proposed to achieve a high density. Among them, bit-pattern media recording (BPMR) is a promising technology. In BPMR systems, data are stored on magnetic islands. Therefore, high densities can be achieved by reducing the distance between the magnetic islands. Because of the closeness between the magnetic islands, the readback signal is distorted by two-dimensional (2D) interference, which includes the intersymbol interference according to the down-track direction and the intertrack interference according to the cross-track direction. A simple and effective serial detection algorithm was recently proposed to mitigate the 2D interference. However, serial detection utilizes the hard output in inner detection, and this degrades the serial detection performance. To resolve this problem, a subsequent study used feedback to estimate the noise and used this noise signal to create a soft output for inner detection. Following up, in this paper we propose a model that utilizes a neural network for noise prediction. The proposed neural network-based model and the model with the feedback line were compared in terms of bit error rate (BER). The results show that the proposed model achieves a gain of approximately 1 dB at a BER of 10−6.

In this paper, an operational Dst index prediction model is developed by combining empirical and Artificial Neural Network (ANN) models. ANN algorithms are widely used to predict space weather conditions. While they require a large amount of data for machine learning, large-scale geomagnetic storms have not occurred sufficiently for the last 20 years, Advanced Composition Explorer (ACE) and Deep Space Climate Observatory (DSCOVR) mission operation period. Conversely, the empirical models are based on numerical equations derived from human intuition and are therefore applicable to extrapolate for large storms. In this study, we distinguish between Coronal Mass Ejection (CME) driven and Corotating Interaction Region (CIR) driven storms, estimate the minimum Dst values, and derive an equation for describing the recovery phase. The combined Korea Astronomy and Space Science Institute (KASI) Dst Prediction (KDP) model achieved better performance contrasted to ANN model only. This model could be used practically for space weather operation by extending prediction time to 24 h and updating the model output every hour.

Because of the physical and engineering problems of conventional magnetic data storage systems, bit-patterned media recording (BPMR) is expected to be a promising technology for extending the storage density to beyond 1 Tb /in2. To increase the storage density in BPMR systems, the separation between islands in both down- and cross-track directions must be reduced, this reduction results in two-dimensional interference from neighboring symbols in those directions, which is a major performance degradation factor in BPMR. Herein, we propose an iterative signal detection scheme between a Viterbi detector and a multilayer perceptron to improve the performance of a BPMR system. In the proposed signal detection scheme, we use the modified output of a multilayer perceptron as a priori information to improve equalization and extrinsic information to decrease the effect of intertrack interference.

Zero-shot multi-speaker Text-to-Speech (TTS) generates target speaker voices given an input text and the corresponding speaker embedding. In this work, we investigate the effectiveness of the TTS reconstruction objective to improve representation learning for speaker verification. We jointly trained end-to-end Tacotron 2 TTS and speaker embedding networks in a self-supervised fashion. We hypothesize that the embeddings will contain minimal phonetic information since the TTS decoder will obtain that information from the textual input. TTS reconstruction can also be combined with speaker classification to enhance these embeddings further. Once trained, the speaker encoder computes representations for the speaker verification task, while the rest of the TTS blocks are discarded. We investigated training TTS from either manual or ASR-generated transcripts. The latter allows us to train embeddings on datasets without manual transcripts. We compared ASR transcripts and Kaldi phone alignments as TTS inputs, showing that the latter performed better due to their finer resolution. Unsupervised TTS embeddings improved EER by 2.06\% absolute with regard to i-vectors for the LibriTTS dataset. TTS with speaker classification loss improved EER by 0.28\% and 0.73\% absolutely from a model using only speaker classification loss in LibriTTS and Voxceleb1 respectively.

With the development of 5G technology, programs are gradually moving to cloud services. This leads to an increasing demand for storage. In the field of high-density data storage, bit-pattern media recording (BPMR) is considered a promising approach, as it can expand the data density to 4 Tb/in2. However, in high-density BPMR, bits or magnetic islands are very close to each other, leading to significant intertrack interference (ITI) from the cross-track direction and intersymbol interference (ISI) from the down-track direction. To minimize two-dimensional interference, including ITI and ISI, the serial detector method has been highly effective. However, in this method, the signal at the output of the first decoder is still a hard output. Therefore, we suggest methods to convert the output of the first detector into a soft output. Additionally, we have developed a new form of generalized partial response target to overcome the track mis-registration. The results show that our proposed methods apparently improve bit error rate performance.

This study investigated the effects of target size and error rate on cognitive demand during augmented reality (AR) interactions. In a repeated-measures laboratory study, twenty participants performed two AR tasks (omni-directional pointing and cube placing) with different target sizes and error rates. During the AR tasks, we measured cerebral oxygenation using functional near-infrared spectroscopy (fNIRS), perceived workload using the NASA-TLX questionnaire, stress using the Short Stress State Questionnaire, and task performance (task completion time). The results showed that the AR tasks with more interaction errors increased cerebral oxygenation, perceived workload, and task completion time while the target size significantly affected physical demand and task completion time. These results suggest that appropriate target sizes and low system errors may reduce potential cognitive demand in AR interactions.

OBJECTIVE. A fundamental goal of the auditory system is to parse the auditory environment into distinct perceptual representations. Auditory perception is mediated by the ventral auditory pathway, which includes the ventrolateral prefrontal cortex (vlPFC). Because large-scale recordings of auditory signals are quite rare, the spatiotemporal resolution of the neuronal code that underlies vlPFC’s contribution to auditory perception has not been fully elucidated. Therefore, we developed a modular, chronic, high-resolution, multi-electrode array system with long-term viability in order to identify the information that could be decoded from μECoG vlPFC signals. APPROACH. We molded three separate μECoG arrays into one and implanted this system in a non-human primate. A custom 3D-printed titanium chamber was mounted on the left hemisphere. The molded 294-contact μECoG array was implanted subdurally over the vlPFC. μECoG activity was recorded while the monkey participated in a ‘hearing-in-noise’ task in which they reported hearing a ‘target’ vocalization from a background ‘chorus’ of vocalizations. We titrated task difficulty by varying the sound level of the target vocalization, relative to the chorus (target-to-chorus ratio, TCr). MAIN RESULTS. We decoded the TCr and the monkey’s behavioral choices from the μECoG signal. We analyzed decoding accuracy as a function of number of electrodes, spatial resolution, and time from implantation. Over a one-year period, we found significant decoding with individual electrodes that increased significantly as we decoded simultaneously more electrodes. Further, we found that the decoding for behavioral choice was better than the decoding of TCr. Finally, because the decoding accuracy of individual electrodes varied on a day-by-day basis, electrode arrays with high channel counts ensure robust decoding in the long term. SIGNIFICANCE. Our results demonstrate the utility of high-resolution and high-channel-count, chronic μECoG recording. We developed a surface electrode array that can be scaled to cover larger cortical areas without increasing the chamber footprint.

A new system-level RF desense characterization technique is investigated to assess platform noise in mobile computer systems. The technique uses a highly tuned antenna probe that provides 8-dB higher noise detection sensitivity in Wi-Fi band of interest than conventional near-field probe technique. This characterization technique is validated with mobile PC system and demonstrates accurate spatial RFI noise distribution. In addition, the low-RFI regions are identified and show improved Wi-Fi throughput performance. The presented desense characterization technique can be used to improve radio system performance and mitigate potential RFI risk in mobile computer system applications.

This paper presents Penny, a compiler-directed resilience scheme for protecting GPU register files (RF) against soft errors. Penny replaces the conventional error correction code (ECC) based RF protection by using less expensive error detection code (EDC) along with idempotence based recovery. Compared to the ECC protection, Penny can achieve either the same level of RF resilience yet with significantly lower hardware costs or stronger resilience using the same ECC due to its ability to detect multi-bit errors when it is used solely for detection. In particular, to address the lack of store buffers in GPUs, which causes both checkpoint storage overwriting and the high cost of checkpointing stores, Penny provides several compiler optimizations such as storage coloring and checkpoint pruning. Across 25 benchmarks, Penny causes only ≈3% run-time overhead on average.

The severity of inter-symbol interference (ISI) and inter-track interference (ITI) effects caused by the reduction of the spacing between bit islands in the along- and across-track directions to acquire ultra-high areal densities (ADs) is the main problem in bit-patterned magnetic recording (BPMR) systems. We call these two effects as two-dimensional (2D) interference, which can degrade the overall system performance. Practically, a one-dimensional (1D) modulation code, e.g., run-length limited (RLL) code, is essentially utilized to prevent the severe ISI effect. The use of 2D detectors is also one of the important ways to cope with the severe ITI; however, its complexity is quite high. To avoid the severe ITI effect and complex 2D detectors; therefore, we propose to apply the 1D RLL code as the 2D modulation code. Then, the 2D modulation encoding constraint is employed to reduce the number of states and branches in the trellis of the proposed modified 2D Viterbi detector. Simulation results reveal that the proposed detector not only delivers a lower complexity but also provides superior bit error rate gain, especially at ultra-high ADs and/or large location fluctuation.