A P300-Based Hangul Input System with a Hierarchical Stimulus Presentation Paradigm

Deep Convolutional Neural Networks (CNNs) perform billions of operations for classifying a single input. To reduce these computations, this paper offers a solution that leverages a combination of runtime information and the algorithmic structure of CNNs. Specifically, in numerous modern CNNs, the outputs of compute-heavy convolution operations are fed to activation units that output zero if their input is negative. By exploiting this unique algorithmic property, we propose a predictive early activation technique, dubbed SnaPEA. This technique cuts the computation of convolution operations short if it determines that the output will be negative. SnaPEA can operate in two distinct modes, exact and predictive. In the exact mode, with no loss in classification accuracy, SnaPEA statically re-orders the weights based on their signs and periodically performs a single-bit sign check on the partial sum. Once the partial sum drops below zero, the rest of computations can simply be ignored, since the output value will be zero in any case. In the predictive mode, which trades the classification accuracy for larger savings, SnaPEA speculatively cuts the computation short even earlier than the exact mode. To control the accuracy, we develop a multi-variable optimization algorithm that thresholds the degree of speculation. As such, the proposed algorithm exposes a knob to gracefully navigate the trade-offs between the classification accuracy and computation reduction. Compared to a state-of-the-art CNN accelerator, SnaPEA in the exact mode, yields, on average, 28% speedup and 16% energy reduction in various modern CNNs without affecting their classification accuracy. With 3% loss in classification accuracy, on average, 67.8% of the convolutional layers can operate in the predictive mode. The average speedup and energy saving of these layers are 2.02x and 1.89x, respectively. The benefits grow to a maximum of 3.59x speedup and 3.14x energy reduction. Compared to static pruning approaches, which are complimentary to the dynamic approach of SnaPEA, our proposed technique offers up to 63% speedup and 49% energy reduction across the convolution layers with no loss in classification accuracy.

Motor imagery classification in electroencephalography (EEG)-based brain-computer interface (BCI) systems is an important research area. To simplify the complexity of the classification, selected power bands and electrode channels have been widely used to extract and select features from raw EEG signals, but there is still a loss in classification accuracy in the state-of-the-art approaches. To solve this problem, we propose a discriminative feature extraction algorithm based on power bands with principle component analysis (PCA). First, the raw EEG signals from the motor cortex area were filtered using a bandpass filter with μ and β bands. This research considered the power bands within a 0.4 second epoch to select the optimal feature space region. Next, the total feature dimensions were reduced by PCA and transformed into a final feature vector set. The selected features were classified by applying a support vector machine (SVM). The proposed method was compared with a state-of-art power band feature and shown to improve classification accuracy.

In this work we present key elements of DeepChip, a framework that bridges recent trends in machine learning with applicable forward propagation on resource-constrained devices. Main objective of this work is to reduce compute and memory requirements by removing redundancy from neural networks. DeepChip features a flexible quantizer to reduce the bit width of activations to 8-bit fixed-point and weights to an asymmetric ternary representation. In combination with novel algorithms and data compression we leverage reduced precision and sparsity for efficient forward propagation on a wide range of processor architectures. We validate our approach on a set of different convolutional neural networks and datasets: ConvNet on SVHN, ResNet-44 on CIFAR10 and AlexNet on ImageNet. Compared to single-precision floating point, memory requirements can be compressed by a factor of 43, 22 and 10 and computations accelerated by a factor of 5.2, 2.8 and 2.0 on a mobile processor without a loss in classification accuracy. DeepChip allows trading accuracy for efficiency, and for instance tolerating about 2% loss in classification accuracy further reduces memory requirements by a factor of 88, 29 and 13, and speeds up computations by a factor of 6.0, 4.3 and 5.0. Code related to this paper is available at: https://github.com/UniHD-CEG/ECML2018.

With the development of high spatial resolution satellites, such as IKONOS, we are able to make more use of spatial features to improve classification accuracy, especially for land-use mapping. In this paper, we compared the texture spectrum (TS) approach with 4 different schemes of gray level reduction, which are min-max linear compression (LC), gray level binning (BN), histogram equalization (HE) and piece-wise nonlinear compression (PC), on the panchromatic band of the sampled IKONOS CARTERRA imagery in terms of overall accuracy and kappa coefficient. The classification algorithm is on the basis of a frequency-based contextual approach that utilizes neighboring information within a particular window. The ranking of 5 different approaches is TS, PC, HE, BN and LC. TS achieves an overall classification accuracy of 72% at a window size of 53 while PC reaches an overall accuracy of 68% at a window size of 35 and LC reaches its highest accuracy of 65% at a window size of 29. TS approach needs a relatively larger window size to achieve higher classification accuracy while the other 4 gray level reduction approaches reach their higher accuracy at smaller window sizes. Within 9 land-use classes, a relatively large window size is more applicable to classify heterogeneous land-use types or areas having first order trend while smaller window sizes work better for more homogeneous land-use types or area with second order effect. The TS approach was found to be more capable of capturing first order trends while the other 4 approaches were more sensitive to local gray level variations thus better capturing the second order effect.

Physical principles of brain–computer interfaces and their applications for rehabilitation, robotics and control of human brain states

Most of the electroencephalography (EEG) based emotion recognition systems rely on single stimulus to evoke emotions. EEG data is mostly recorded with higher number of electrodes that can lead to data redundancy and longer experimental setup time. The question “whether the configuration with lesser number of electrodes is common amongst different stimuli presentation paradigms” remains unanswered. There are publicly available datasets for EEG based human emotional states recognition. Since this work is focused towards classifying emotions while subjects are experiencing different stimuli, therefore we need to perform new experiments. Keeping aforementioned issues in consideration, this work presents a novel experimental study that records EEG data for three different human emotional states evoked with four different stimuli presentation paradigms. A methodology based on iterative Genetic Algorithm in combination with majority voting has been used to achieve configuration with reduced number of EEG electrodes keeping in consideration minimum loss of classification accuracy. The results obtained are comparable with recent studies. Stimulus independent configurations with lesser number of electrodes lead towards low computational complexity as well as reduced set up time for future EEG based smart systems for emotions recognition

Motor intention in the posterior parietal cortex

Single-trial motor imagery classification is a crucial aspect of brain–computer applications. Therefore, it is necessary to extract and discriminate signal features involving motor imagery movements. Riemannian geometry-based feature extraction methods are effective when designing these types of motor-imagery-based brain–computer interface applications. In the field of information theory, Riemannian geometry is mainly used with covariance matrices. Accordingly, investigations showed that if the method is used after the execution of the filterbank approach, the covariance matrix preserves the frequency and spatial information of the signal. Deep-learning methods are superior when the data availability is abundant and while there is a large number of features. The purpose of this study is to a) show how to use a single deep-learning-based classifier in conjunction with BCI (brain–computer interface) applications with the CSP (common spatial features) and the Riemannian geometry feature extraction methods in BCI applications and to b) describe one of the wrapper feature-selection algorithms, referred to as the particle swarm optimization, in combination with a decision tree algorithm. In this work, the CSP method was used for a multiclass case by using only one classifier. Additionally, a combination of power spectrum density features with covariance matrices mapped onto the tangent space of a Riemannian manifold was used. Furthermore, the particle swarm optimization method was implied to ease the training by penalizing bad features, and the moving windows method was used for augmentation. After empirical study, the convolutional neural network was adopted to classify the pre-processed data. Our proposed method improved the classification accuracy for several subjects that comprised the well-known BCI competition IV 2a dataset.

Pattern recognition involves identifying a new instance as a member of a class based on the features extracted from problem space. Not all features contribute to increase in accuracy of classification. Wrapper based methodology selects an optimal subset of reduced features with no much loss in classification accuracy. Inspired by the bio evolutionary behaviour of flowers in optimal reproduction, a wrapper based optimal feature selection methodology is devised. In this paper, a Binary Flower Pollination Algorithm (BFPA) is proposed to address the combinatorial problem of feature selection. Support Vector machine (SVM) is used as a subset evaluator to calculate the fitness function to be optimized. Proposed algorithm is evaluated using 5 datasets from UCI repository. Switching probability is varied to achieve a fair tradeoff between exploration and exploitation. Feature subset chosen by the proposed method serves as input to 6 traditional classifiers to analyse the success of the algorithm. Experimental results are also compared with existing literature and observed that the proposed methodology identifies feature subset with reduced features and improved accuracy.

Brain–computer interfaces (BCIs), also known as brain–machine interfaces (BMIs), translate brain activity into new outputs that replace, restore, enhance, supplement or improve natural brain outputs. BCI research and development has grown rapidly for the past two decades. It is beginning to provide useful communication and control capacities to people with severe neuromuscular disabilities; and it is expanding into new areas such as neurorehabilitation that may greatly increase its clinical impact. At the same time, significant challenges remain, particularly in regard to translating laboratory advances into clinical use.

Chapter 10 On the Use of Brain–Computer Interfaces Outside Scientific Laboratories: Toward an Application in Domotic Environments

Steady State Visual Evoked Potential (SSVEP) has been commonly adopted in Brain Computer Interface (BCI) applications. For wearable BCI applications, several aspects of SSVEP-based BCI systems, such as speed, subject variability, and accurate target detection, are under ongoing research investigations. Up to date, Canonical Correlation Analysis (CCA) has been considered the state-of-the-art feature extraction method for SSVEP-based BCI systems. Nevertheless, although CCA outperforms traditional SSVEP detection methods, such as Power Spectral Density Analysis (PSDA), achieving high accuracies when detecting target frequencies is still a challenging task due to user variation and physiological changes in the human body. In this paper, we investigate an SSVEP-based BCI application using wireless EEG recording and an Android tablet-based user interface. We propose a fusion of CCA and PSDA solutions at the score level by dividing their score space into multiple partitions, and extract and combine their complementary discriminative information to minimize the detection error in a linear fashion. We investigated transforming the fusion score space to lower dimensions with the purpose of alleviating redundancy. As such, we employed Principal Component Analysis (PCA), and Linear Discriminant Analysis (LDA). Our experimental results demonstrated that our proposed score fusion method is effective in reducing the effect of noise and non-stationary elements in EEG dynamics. Average detection accuracies improved from 63% for CCA to 72% for fusion+PCA and further improved to 98% for fusion+LDA.

Cortical activation and BCI performance during brief tactile imagery: A comparative study with motor imagery

This paper introduces an automated method for estimating sex from the lower and upper limbs based on diaphyseal CSG properties. The proposed method was developed and evaluated using 389 femurs, 412 tibias, and 404 humeri of adult individuals from a modern Greek reference sample, the Athens Collection. The skeletal properties, which were extracted with the CSG-Toolkit, were analyzed with step-wise DFA (evaluated with LOOCV) and subsequently with RBF kernel SVM supervised learning. SVM cross-validation was based on a 20-fold stratified random sample splitting as well as a chronological split based on year of birth to further assess the effect of secular change in sex estimation capacity. Maximum cross-validated classification accuracy from step-wise DFA reached 94.8% for the femur, 94.7% for the tibia, and 97.3% for the humerus, whereas SVM cross-validated results were similar although slightly lower, mainly due to the more strict cross-validation scheme. Our results suggest that the proposed sex estimation method is reasonably robust to secular change, since there was limited loss in classification accuracy between different chronological groups, despite the presence of secular change in stature of the Greek population during the examined period. The proposed method has been implemented as a function for the GNU Octave environment, named estimate_sex, which comprises a self-intuitive graphical user interface for facilitating sex estimation and is freely available under a suitable license.

New instruments currently in design for Earth observational remote sensing from space offer promising advancements in the use of remote sensing for soil studies. Due to the enormous quantities of data generated by these high spectral resolution sensors, efficient use will only be feasible if new algorithms are developed to compress the data while maintaining information content. A spectral band selection algorithm is proposed to identify the important spectral bands for compression and classification of soil reflectance data. This algorithm was developed from the shape dominancy concept of Karhunen‐Loeve based optimal features. The high dimensional raw data were first transformed into much lower dimensionality by the derived spectral features. Canonical analysis was then used to transform the data under the maximal separability criterion into their final signal space where the data classification was performed. Soil data sets with and without stratification by soil order and climatic moisture zone were used to test the algorithm. The probabilities of correct classification of soil organic matter content using Landsat Multispectral Scanner (MSS) bands, Thematic Mapper (TM) bands, and the bands identified using the new band selection algorithm, were calculated and compared. The algorithm was successful in finding important spectral bands for soil organic matter content classification. The probabilities of correct classification obtained by using these bands were 0.850 and 0.883 for unstratified data sets, as compared to 0.600 to 0.640 for MSS bands and 0.640 to 0.653 for TM bands. Probabilities of correct classification for climate‐stratified data ranged from 0.910 to 0.980 using the calculated bands. The overall data compression ratio achieved using the algorithm was greater than 10 with no loss in classification accuracy.