Number Of Input Channels Research Articles

ICA’s bug: How ghost ICs emerge from effective rank deficiency caused by EEG electrode interpolation and incorrect re-referencing

Independent component analysis (ICA) has been widely used for electroencephalography (EEG) analyses. However, ICA performance relies on several crucial assumptions about the data. Here, we focus on the granularity of data rank, i.e., the number of linearly independent EEG channels. When the data are rank-full (i.e., all channels are independent), ICA produces as many independent components (ICs) as the number of input channels (rank-full decomposition). However, when the input data are rank-deficient, as is the case with bridged or interpolated electrodes, ICA produces the same number of ICs as the data rank (forced rank deficiency decomposition), introducing undesired ghost ICs and indicating a bug in ICA. We demonstrated that the ghost ICs have white noise properties, in both time and frequency domains, while maintaining surprisingly typical scalp topographies, and can therefore be easily missed by EEG researchers and affect findings in unknown ways. This problem occurs when the minimum eigenvalue λmin of the input data is smaller than a certain threshold, leading to matrix inversion failure as if the rank-deficient inversion was forced, even if the data rank is cleanly deficient by one. We defined this problem as the effective rank deficiency. Using sound file mixing simulations, we first demonstrated the effective rank deficiency problem and determined that the critical threshold for λmin is 10−7 in the given situation. Second, we used empirical EEG data to show how two preprocessing stages, re-referencing to average without including the initial reference and non-linear electrode interpolation, caused this forced rank deficiency problem. Finally, we showed that the effective rank deficiency problem can be solved by using the identified threshold (λmin = 10−7) and the correct re-referencing procedure described herein. The former ensures the achievement of effective rank-full decomposition by properly reducing the input data rank, and the latter allows avoidance of a widely practiced incorrect re-referencing approach. Based on the current literature, we discuss the ambiguous status of the initial reference electrode when re-referencing. We have made our data and code available to facilitate the implementation of our recommendations by the EEG community.

Reduced volume of diabetic pancreatic islets in rodents detected by synchrotron X-ray phase-contrast microtomography and deep learning network

The pancreatic islet is a highly structured micro-organ that produces insulin in response to rising blood glucose. Here we develop a label-free and automatic imaging approach to visualize the islets in situ in diabetic rodents by the synchrotron radiation X-ray phase-contrast microtomography (SRμCT) at the ID17 station of the European Synchrotron Radiation Facility. The large-size images (3.2mm×15.97mm) were acquired in the pancreas in STZ-treated mice and diabetic GK rats. Each pancreas was dissected by 3000 reconstructed images. The image datasets were further analysed by a self-developed deep learning method, AA-Net. All islets in the pancreas were segmented and visualized by the three-dimension (3D) reconstruction. After quantifying the volumes of the islets, we found that the number of larger islets (=>1500 μm3) was reduced by 2-fold (wt 1004±94 vs GK 419±122, P<0.001) in chronically developed diabetic GK rat, while in STZ-treated diabetic mouse the large islets were decreased by half (189±33 vs 90±29, P<0.001) compared to the untreated mice. Our study provides a label-free tool for detecting and quantifying pancreatic islets in situ. It implies the possibility of monitoring the state of pancreatic islets in vivo diabetes without labelling.

Grouped Pointwise Convolutions Reduce Parameters in Convolutional Neural Networks

In DCNNs, the number of parameters in pointwise convolutions rapidly grows due to the multiplication of the number of filters by the number of input channels that come from the previous layer. Our proposal makes pointwise convolutions parameter efficient via grouping filters into parallel branches or groups, where each branch processes a fraction of the input channels. However, by doing so, the learning capability of the DCNN is degraded. To avoid this effect, we suggest interleaving the output of filters from different branches at intermediate layers of consecutive pointwise convolutions. We applied our improvement to the EfficientNet, DenseNet-BC L100, MobileNet and MobileNet V3 Large architectures. We trained these architectures with the CIFAR-10, CIFAR-100, Cropped-PlantDoc and The Oxford-IIIT Pet datasets. When training from scratch, we obtained similar test accuracies to the original EfficientNet and MobileNet V3 Large architectures while saving up to 90% of the parameters and 63% of the flops.

Efficient emotion recognition using hyperdimensional computing with combinatorial channel encoding and cellular automata

In this paper, a hardware-optimized approach to emotion recognition based on the efficient brain-inspired hyperdimensional computing (HDC) paradigm is proposed. Emotion recognition provides valuable information for human–computer interactions; however, the large number of input channels (> 200) and modalities (> 3 ) involved in emotion recognition are significantly expensive from a memory perspective. To address this, methods for memory reduction and optimization are proposed, including a novel approach that takes advantage of the combinatorial nature of the encoding process, and an elementary cellular automaton. HDC with early sensor fusion is implemented alongside the proposed techniques achieving two-class multi-modal classification accuracies of > 76% for valence and > 73% for arousal on the multi-modal AMIGOS and DEAP data sets, almost always better than state of the art. The required vector storage is seamlessly reduced by 98% and the frequency of vector requests by at least 1/5. The results demonstrate the potential of efficient hyperdimensional computing for low-power, multi-channeled emotion recognition tasks.

A Novel and Automated Approach to Classify Radiation Induced Lung Tissue Damage on CT Scans.

Simple SummaryRadiation-induced lung damage (RILD) is a common side effect of treating lung cancer with radiotherapy (RT). RILD is visible on CT imaging, and its radiological appearance can vary dramatically from patient to patient as well as across different sub-regions of the lung and treatment volumes. A classification system for RILD able to differentiate radiological damage on a local level would allow us to better understand the underlying patterns of RILD, see how they change over time post irradiation, and link it with clinical outcomes. In this work we propose a five-class morphological lung tissue classification system that can describe parenchymal tissue changes at the voxel level. The classifier was implemented in a fully automated manner using an optimised deep-learning method, then trained and tested using data acquired through a multi-centre clinical trial. The proposed method performed well on an unseen testing dataset. The automated segmentation achieved considerable overlap with manual segmentations (ranging between 26% and 98% for the five classes) and was graded as acceptable by a clinical expert in 88% of cases. This demonstrates it to be suitable for application on a large dataset to help uncover different patterns of changes in the population.Radiation-induced lung damage (RILD) is a common side effect of radiotherapy (RT). The ability to automatically segment, classify, and quantify different types of lung parenchymal change is essential to uncover underlying patterns of RILD and their evolution over time. A RILD dedicated tissue classification system was developed to describe lung parenchymal tissue changes on a voxel-wise level. The classification system was automated for segmentation of five lung tissue classes on computed tomography (CT) scans that described incrementally increasing tissue density, ranging from normal lung (Class 1) to consolidation (Class 5). For ground truth data generation, we employed a two-stage data annotation approach, akin to active learning. Manual segmentation was used to train a stage one auto-segmentation method. These results were manually refined and used to train the stage two auto-segmentation algorithm. The stage two auto-segmentation algorithm was an ensemble of six 2D Unets using different loss functions and numbers of input channels. The development dataset used in this study consisted of 40 cases, each with a pre-radiotherapy, 3-, 6-, 12-, and 24-month follow-up CT scans (n = 200 CT scans). The method was assessed on a hold-out test dataset of 6 cases (n = 30 CT scans). The global Dice score coefficients (DSC) achieved for each tissue class were: Class (1) 99% and 98%, Class (2) 71% and 44%, Class (3) 56% and 26%, Class (4) 79% and 47%, and Class (5) 96% and 92%, for development and test subsets, respectively. The lowest values for the test subsets were caused by imaging artefacts or reflected subgroups that occurred infrequently and with smaller overall parenchymal volumes. We performed qualitative evaluation on the test dataset presenting manual and auto-segmentation to a blinded independent radiologist to rate them as ‘acceptable’, ‘minor disagreement’ or ‘major disagreement’. The auto-segmentation ratings were similar to the manual segmentation, both having approximately 90% of cases rated as acceptable. The proposed framework for auto-segmentation of different lung tissue classes produces acceptable results in the majority of cases and has the potential to facilitate future large studies of RILD.

Effect of filter sizes on image classification in CNN: a case study on CFIR10 and Fashion-MNIST datasets

&lt;span style="font-size: 9pt; font-family: 'Times New Roman', serif;"&gt;Convolution neural networks (CNN or ConvNet), a deep neural network class inspired by biological processes, are immensely used for image classification or visual imagery. These networks need various parameters or attributes like number of filters, filter size, number of input channels, padding stride and dilation, for doing the required task. In this paper, we focused on the hyperparameter, i.e., filter size. Filter sizes come in various sizes like 3×3, 5×5, and 7×7. We varied the filter sizes and recorded their effects on the models' accuracy. The models' architecture is kept intact and only the filter sizes are varied. This gives a better understanding of the effect of filter sizes on image classification. CIFAR10 and FashionMNIST datasets are used for this study. Experimental results showed the accuracy is inversely proportional to the filter size. The accuracy using 3×3 filters on CIFAR10 and Fashion-MNIST is 73.04% and 93.68%, respectively.&lt;/span&gt;

SILIA: software implementation of a multi-channel, multi-frequency lock-in amplifier for spectroscopy and imaging applications

We describe a software implementation of a multi-channel, multi-frequency Lock-in Amplifier (SILIA) to extract modulated signals from noisy data distributed over multiple channels of arbitrary number and size. This software implementation emulates the functionality of a multi-channel, multi-frequency lock-in amplifier in a post-processing step following data acquisition. Unlike most traditional lock-in amplifiers, SILIA can work with any number of input channels and is especially useful to analyze data distributed over many channels. We demonstrate the versatility and performance for extracting weak signals in spectroscopy and fluorescence microscopy. We also discuss more general applications and exhibit a method to automatically estimate error from a lock-in result.

Efficient and versatile toolbox for analysis of time-tagged measurements

Acquisition and analysis of time-tagged events is a ubiquitous tool in scientific and industrial applications. With increasing time resolution, number of input channels, and acquired events, the amount of data can be overwhelming for standard processing techniques. We developed the Extensible Time-tag Analyzer (ETA), a powerful and versatile, yet easy to use software to efficiently analyze and display time-tagged data. Our tool allows for flexible extraction of correlation from time-tagged data beyond start-stop measurements that were traditionally used. A combination of state diagrams and simple code snippets allows for analysis of arbitrary complexity while keeping computational efficiency high.

A Flexible Multichannel EEG Artifact Identification Processor using Depthwise-Separable Convolutional Neural Networks

This article presents an energy-efficient and flexible multichannel Electroencephalogram (EEG) artifact identification network and its hardware using depthwise and separable convolutional neural networks. EEG signals are recordings of the brain activities. EEG recordings that are not originated from cerebral activities are termed artifacts . Our proposed model does not need expert knowledge for feature extraction or pre-processing of EEG data and has a very efficient architecture implementable on mobile devices. The proposed network can be reconfigured for any number of EEG channel and artifact classes. Experiments were done with the proposed model with the goal of maximizing the identification accuracy while minimizing the weight parameters and required number of operations. Our proposed network achieves 93.14% classification accuracy using an EEG dataset collected by 64-channel BioSemi ActiveTwo headsets, averaged across 17 patients and 10 artifact classes. Our hardware architecture is fully parameterized with number of input channels, filters, depth, and data bit-width. The number of processing engines (PE) in the proposed hardware can vary between 1 to 16, providing different latency, throughput, power, and energy efficiency measurements. We implement our custom hardware architecture on Xilinx FPGA (Artix-7), which on average consumes 1.4 to 4.7 mJ dynamic energy with different PE configurations. Energy consumption is further reduced by 16.7× implementing on application-specified integrated circuit at the post layout level in 65-nm CMOS technology. Our FPGA implementation is 1.7 × to 5.15 × higher in energy efficiency than some previous works. Moreover, our Application-Specified Integrated Circuit implementation is also 8.47 × to 25.79 × higher in energy efficiency compared to previous works. We also demonstrated that the proposed network is reconfigurable to detect artifacts from another EEG dataset collected in our lab by a 14-channel Emotiv EPOC+ headset and achieved 93.5% accuracy for eye blink artifact detection.

THE PARALLEL-PIPELINED IMPLEMENTATION OF THE FRACTAL IMAGE COMPRESSION FOR RECONFIGURABLE COMPUTING SYSTEMS

The developed fractal image compression method, implemented for reconfigurable computing systems is described. The main idea parallel fractal image compression based on parallel execution pairwise comparison of domain and rank blocks. Achievement high performance occurs at the expense of simultaneously comparing maximum number of pairs. Implementation fractal image compression for reconfigurable computing systems has two critical resources, as number of input channels and FPGA Look-up Table (LUT). The main critical resource for fractal image compression is data channels, and implementation this task for reconfigurable computing systems requires parallel-pipeline computations organization replace parallel, preliminarily produced performance reduction parallel computational structure. The main critical resource for fractal image compression is data channels, and implementation this task for reconfigurable computing systems requires parallel-pipeline computations organization replace parallel computations organiation. For using parallel-pipeline computations organization, preliminarily have produce performance reduction parallel computational structure. Each operator has routed to computational structure sequentially (bit by bit) to save computational resources and reduces equipment downtime. Storing iterated functions system coefficients for image encoding has been introduced in data structure, which correlates between corresponding parameters the numbers of rank and domain blocks. Applying this approach for parallel-pipeline programs allows scaling computing structure to plurality programmable logic arrays (FPGAs). Task implementation on the reconfigurable computer system Tertius-2 containing eight FPGAs 15 000 times provides performed acceleration relatively with universal multi-core processor, and 18 – 25 times whit to existing solutions for FPGAs.

Progressive fusion of feature sets for optimal classification of MI signal

Brain-computer interface (BCI) is the most popular research topic to the researchers of neuroprosthetics. The ultimate goal of this research is to develop a communication channel between human brains and external devices. Feature extraction is one of the most crucial steps in this research. Combination of different features may improve the classification performance, but in most of the cases straight forward combinations of different features lead to a very poor result. So, it is necessary to combine the orthogonal features and omit the redundant ones. It is a complex and time-consuming process. We have developed two new algorithms to find optimum sets of features for fusion to obtain best possible classification accuracy for both subject-specific (SS) and subject-independent (SI) cases. Experimental results indicate that our proposed algorithms, in general, improve the classification results irrespective of the different methodological setup of BCI processes such as number of input channels and spatial filter.

Autonomous Process Model Identification using Recurrent Neural Networks and Hyperparameter Optimization

We demonstrate the application of automated machine learning to the problem of identifying dynamic process models using recurrent neural networks (RNNs). The general concept relies on continuous monitoring of input-output data from a plant and the processing of this data by a collection of algorithms. The data is first processed by a collaborative filtering system, which suggests a classification of system dynamics per output channel. The proposed classification and the number of input channels is used to initialize a search over RNN hyperparameters. The search algorithm uses subsets of historical data for training and validation to select the RNN architecture and determine the network parameters according to preselected objectives for balancing model accuracy and model compactness. The proposed approach is demonstrated on a simulated case study for online system identification of a chemical reactor, where the underlying dynamic characteristics of the simulated system are changed during the simulation as the system undergoes a number of disturbances and handles control tasks. Process models for the system in question are obtained via the automated machine learning approach and the models are updated as the system dynamics change. The results show good prediction accuracy of the models throughout the simulation representing changes in system dynamics.

Asymmetric convolution with densely connected networks

Convolutional neural networks are vital to some computer vision tasks, and the densely connected network is a creative architecture among them. In densely connected network, most convolution layer tends to have a much larger number of input channels than output channels, making itself to a funnel shape. We replace the 3 × 3 convolution in the densely connected network with two continuous asymmetric convolutions to make the DenseNet family more diverse. We also proposed a model in which two continuous asymmetric convolutions each outputs half of the output channels and concatenate them as the final output of these layers. Compared with the original densely connected network, our models achieve similar performance on CIFAR-10/100 dataset with fewer parameters and less computational cost.

GENERATING 3D CITY MODELS BASED ON THE SEMANTIC SEGMENTATION OF LIDAR DATA USING CONVOLUTIONAL NEURAL NETWORKS

Abstract. Virtual city models are important for many applications such as urban planning, virtual and augmented reality, disaster management, and gaming. Urban features such as buildings, roads, and trees are essential components of these models and are subject to frequent change and alteration. It is laborious to manually build and update virtual city models, due to a large number of instances and temporal changes on such features. The increase of publicly available spatial data provides an important source for pipelines that automate virtual city model generation. The large quantity of data also opens an opportunity to use Deep Learning (DL) as a technique that minimizes the need for expert domain knowledge. In addition, many Deep Learning models calculations can be parallelized on modern hardware such as graphical processing units, which reduces the computation time substantially.We explore the opportunity of using publicly available data in computing multiple thematic data layers from Digital Surface Models (DSMs) using an automatic pipeline that is powered by a semantic segmentation network. To evaluate this design, we implement our pipeline using multiple Convolutional Neural Networks (CNN) with an encoder-decoder architecture. We produce a variety of two and three-dimensional thematic data. We focus our evaluation on the pipeline’s ability to produce accurate building footprints. In our experiments we vary the depths, the number of input channels and data resolutions of the evaluated networks. Our experiments process public data that is provided by New York City.

Analysis of Radio over Fiber system for mitigating four-wave mixing effect

In this paper, an efficient 8-channel 32Gbps RoF (Radio over Fiber) system incorporating Bessel Filter (8/32 RoF-BF) has been demonstrated to reduce the impact of non-linear transmission effects, specifically Four-Wave Mixing (FWM). The simulation results indicate that the proposed 8/32 RoF-BF system provides an optimum result w.r.t. channel spacing (75 GHz), input source power (0 dBm) and number of input channels (8). In comparison with the existing RoF system, the proposed 8/32 RoF-BF system has been validated analytically and it is found that the performance of the proposed system is in close proximity particularly in FWM sideband power reduction of the order of 4 dBm for the 8-channel 32Gbps RoF system.

ORTHOSEG: A DEEP MULTIMODAL CONVOLUTONAL NEURAL NETWORK ARCHITECTURE FOR SEMANTIC SEGMENTATION OF ORTHOIMAGERY

Abstract. This paper addresses the task of semantic segmentation of orthoimagery using multimodal data e.g. optical RGB, infrared and digital surface model. We propose a deep convolutional neural network architecture termed OrthoSeg for semantic segmentation using multimodal, orthorectified and coregistered data. We also propose a training procedure for supervised training of OrthoSeg. The training procedure complements the inherent architectural characteristics of OrthoSeg for preventing complex co-adaptations of learned features, which may arise due to probable high dimensionality and spatial correlation in multimodal and/or multispectral coregistered data. OrthoSeg consists of parallel encoding networks for independent encoding of multimodal feature maps and a decoder designed for efficiently fusing independently encoded multimodal feature maps. A softmax layer at the end of the network uses the features generated by the decoder for pixel-wise classification. The decoder fuses feature maps from the parallel encoders locally as well as contextually at multiple scales to generate per-pixel feature maps for final pixel-wise classification resulting in segmented output. We experimentally show the merits of OrthoSeg by demonstrating state-of-the-art accuracy on the ISPRS Potsdam 2D Semantic Segmentation dataset. Adaptability is one of the key motivations behind OrthoSeg so that it serves as a useful architectural option for a wide range of problems involving the task of semantic segmentation of coregistered multimodal and/or multispectral imagery. Hence, OrthoSeg is designed to enable independent scaling of parallel encoder networks and decoder network to better match application requirements, such as the number of input channels, the effective field-of-view, and model capacity.

Transmission matrix optimization based on singular value decomposition in strong scattering process

In the last decade, the scattering medium has been gradually attacking attention from researchers. Among the proposed approaches, the transmission matrix (TM) is considered as an effect way to describe the scattering properties which relate to input optical and output optical fields. However, the acquired transmission matrix and its eigenvalues strongly depend on the experimental conditions, such as the numbers of input channels (limited numerical aperture and illumination area, or the pixel number of the spatial light modulator) and output channels. In other words, the actual transmission matrix of the scattering medium is the acquired transmission matrix with infinite numbers of the input and output channels. We propose an approach to obtaining the actual matrix by evaluating its eigenvalues. First, the matrix is expressed by the singular value decomposition to obtain its inverse matrix. Then first level optimization is to dispose of some extreme singular values to remove the ill-conditioned problem of the matrix, and then, as a second level optimization, the genetic algorithm is to remove the eigenvalues which have the negative contributions to the intensity of the selected focal point. Our experiments show that the gray value of the intensity and the signal-to-noise ratio (SNR) of the focal point after employing the phase pattern are 129 and 7.54, respectively. After the first level optimization, the gray value of the intensity and the SNR rise to 172 and 9.73, respectively. Then, they reach to 192 and 10.29, respectively, after adopting the genetic algorithm. After the second level optimizations, the intensity at the focal point increases 48.8% compared with the case with just the optimized phase pattern from the acquired TM, and the SNR increases by nearly 36.5%. The reason behind the increase of the intensity after the optimizations, we believe, is that the transmission matrix of the scattering medium reaches its actual matrix in certain conditions. The proposed approach opens the way to obtaining the actual transmission matrix by mathematic optimizations without increasing the experimental levels.

WeedNet: Dense Semantic Weed Classification Using Multispectral Images and MAV for Smart Farming

Selective weed treatment is a critical step in autonomous crop management as related to crop health and yield. However, a key challenge is reliable, and accurate weed detection to minimize damage to surrounding plants. In this paper, we present an approach for dense semantic weed classification with multispectral images collected by a micro aerial vehicle (MAV). We use the recently developed encoder-decoder cascaded Convolutional Neural Network (CNN), Segnet, that infers dense semantic classes while allowing any number of input image channels and class balancing with our sugar beet and weed datasets. To obtain training datasets, we established an experimental field with varying herbicide levels resulting in field plots containing only either crop or weed, enabling us to use the Normalized Difference Vegetation Index (NDVI) as a distinguishable feature for automatic ground truth generation. We train 6 models with different numbers of input channels and condition (fine-tune) it to achieve about 0.8 F1-score and 0.78 Area Under the Curve (AUC) classification metrics. For model deployment, an embedded GPU system (Jetson TX2) is tested for MAV integration. Dataset used in this paper is released to support the community and future work.

Development of the Graphical User Interface of Mobile Application for Remote Control Using Wireless Communications

The single most important concept to master when designing mobile device interfaces is context. The context in which an application is used and the context of how information is input are both key issues; each must be understood before a well-crafted design may be implemented. When these two notions of context are explored, it becomes clear that designing for a mobile device can lead to a solution that is worlds different than its desktop equivalent. In this paper, we developed a graphical user interface of mobile application to control a digital audio system remotely. Most significantly, this remote control mobile application for the control of digital audio must be able to control many channels simultaneously. The number of input channels is 4 and the number of output is 8. Also, it has a lot of graphical components and has many kinds of components that are controlled by the mobile application. It creates an experience especially for many users of digital audio, allowing the user to interact with the software for digital audio operation in a natural and intuitive way. As a result, we implemented the mobile application interface of digital audio to provide a user-friendly experience. An efficient memory management is also necessary in order to execute the operation for the remote control and hard cording is needed on the mobile application. To solve that problem, a GridLayout was used. As a result, the screen size is not problem in operating the mobile application on android.

Investigation of four wave mixing effect with different number of input channels at various channel spacing

In this paper the design, implementation and performance analysis of four wave mixing (FWM) in optical communication system for different number of input channels is presented using various values of channel spacing. Here, all the input channels have been spaced evenly at various values like 6.25GHz, 12.5GHz, 25GHz, 40GHz, 50GHz with the different number of channels at the input i.e. with 2, 4, 6, 8, 12 input channels. The simulation results reveal that the four wave mixing is minimum when the channel spacing is maximum i.e. 50GHz with minimum number of channels i.e. 2 input channels. It is observed that on increasing the channel spacing, the interference between the input frequencies decreases and hence the four wave mixing also decreases. Also, on increasing the number of input channels/users, the interference between the input frequencies increases and thus, the four wave mixing also increases.