nips nips2007 nips2007-74 knowledge-graph by maker-knowledge-mining
Source: pdf
Author: Pierre Ferrez, José Millán
Abstract: Brain-computer interfaces (BCIs), as any other interaction modality based on physiological signals and body channels (e.g., muscular activity, speech and gestures), are prone to errors in the recognition of subject’s intent. An elegant approach to improve the accuracy of BCIs consists in a verification procedure directly based on the presence of error-related potentials (ErrP) in the EEG recorded right after the occurrence of an error. Six healthy volunteer subjects with no prior BCI experience participated in a new human-robot interaction experiment where they were asked to mentally move a cursor towards a target that can be reached within a few steps using motor imagination. This experiment confirms the previously reported presence of a new kind of ErrP. These “Interaction ErrP” exhibit a first sharp negative peak followed by a positive peak and a second broader negative peak (∼290, ∼350 and ∼470 ms after the feedback, respectively). But in order to exploit these ErrP we need to detect them in each single trial using a short window following the feedback associated to the response of the classifier embedded in the BCI. We have achieved an average recognition rate of correct and erroneous single trials of 81.8% and 76.2%, respectively. Furthermore, we have achieved an average recognition rate of the subject’s intent while trying to mentally drive the cursor of 73.1%. These results show that it’s possible to simultaneously extract useful information for mental control to operate a brain-actuated device as well as cognitive states such as error potentials to improve the quality of the braincomputer interaction. Finally, using a well-known inverse model (sLORETA), we show that the main focus of activity at the occurrence of the ErrP are, as expected, in the pre-supplementary motor area and in the anterior cingulate cortex. 1
Reference: text
sentIndex sentText sentNum sentScore
1 ch ∗ Abstract Brain-computer interfaces (BCIs), as any other interaction modality based on physiological signals and body channels (e. [sent-9, score-0.222]
2 Six healthy volunteer subjects with no prior BCI experience participated in a new human-robot interaction experiment where they were asked to mentally move a cursor towards a target that can be reached within a few steps using motor imagination. [sent-13, score-0.85]
3 These “Interaction ErrP” exhibit a first sharp negative peak followed by a positive peak and a second broader negative peak (∼290, ∼350 and ∼470 ms after the feedback, respectively). [sent-15, score-0.47]
4 But in order to exploit these ErrP we need to detect them in each single trial using a short window following the feedback associated to the response of the classifier embedded in the BCI. [sent-16, score-0.246]
5 We have achieved an average recognition rate of correct and erroneous single trials of 81. [sent-17, score-0.312]
6 Furthermore, we have achieved an average recognition rate of the subject’s intent while trying to mentally drive the cursor of 73. [sent-20, score-0.333]
7 These results show that it’s possible to simultaneously extract useful information for mental control to operate a brain-actuated device as well as cognitive states such as error potentials to improve the quality of the braincomputer interaction. [sent-22, score-0.316]
8 Finally, using a well-known inverse model (sLORETA), we show that the main focus of activity at the occurrence of the ErrP are, as expected, in the pre-supplementary motor area and in the anterior cingulate cortex. [sent-23, score-0.601]
9 In particular, [6] recently reported the presence of a new kind of error potentials (ErrP) elicited by erroneous feedback provided by a BCI during the recognition of the subject’s intent. [sent-36, score-0.527]
10 In this study subjects were asked to reach a target by sending repetitive manual commands to pass over several steps. [sent-37, score-0.385]
11 The system was executing commands with an 80% accuracy, so that at each step there was a 20% probability that the system delivered an erroneous feedback. [sent-38, score-0.264]
12 The main components of these “Interaction ErrP” are a negative peak 250 ms after the feedback, a positive peak 320 ms after the feedback and a second broader negative peak 450 ms after the feedback. [sent-39, score-0.901]
13 To exploit these ErrP for BCIs, it is mandatory to detect them no more in grand averages but in each single trial using a short window following the feedback associated to the response of the BCI. [sent-40, score-0.292]
14 The reported average recognition rates of correct and erroneous single trials are 83. [sent-41, score-0.312]
15 However, it is to note that in order to isolate the issue of the recognition of ErrP out of the more difficult and general problem of a whole BCI where erroneous feedback can be due to nonoptimal performance of both the interface (i. [sent-45, score-0.37]
16 , the classifier embedded into the interface) and the user himself, the subject delivered commands manually. [sent-47, score-0.252]
17 The key issue now is to investigate whether subjects also show ErrP while already engaged in tasks that require a high level of concentration such as motor imagination, and no more in easy tasks such as pressing a key. [sent-48, score-0.373]
18 Subjects don’t deliver manual commands anymore, but are focussing on motor imagination tasks to reach targets randomly selected by the system. [sent-50, score-0.406]
19 1% accuracy in the recognition of the subject’s intent during mental control of the BCI. [sent-53, score-0.244]
20 This confirms the fact that EEG conveys simultaneously information from which we can derive mental commands as well as information about cognitive states and shows that both can be sufficiently well recognized in each single trials to provide the subject with an improved brain-computer interaction. [sent-54, score-0.52]
21 (1) The target (blue) appears 2 steps on the left side of the cursor (green). [sent-57, score-0.359]
22 (2) The subject is imagining a movement of his/her left hand and the cursor moves 1 step to the left. [sent-58, score-0.46]
23 (3) The subject still focuses on his/her left hand, but the system moves the cursor in the wrong direction. [sent-59, score-0.426]
24 (6) A new target (red) appears 3 steps on the right side of the cursor, the subject will now imagine a movement of his/her right foot. [sent-62, score-0.311]
25 The system moved the cursor with an error rate of 20%; i. [sent-63, score-0.278]
26 2 Experimental setup The first step to integrate ErrP detection in a BCI is to design a protocol where the subject is focussing on a mental task for device control and on the feedback delivered by the BCI for ErrP 2 detection. [sent-66, score-0.588]
27 To test the ability of BCI users to concentrate simultaneously on a mental task and to be aware of the BCI feedback at each single trial, we have simulated a human-robot interaction task where the subject has to bring the robot to targets 2 or 3 steps either to the left or to the right. [sent-67, score-0.561]
28 This virtual interaction is implemented by means of a green square cursor that can appear on any of 20 positions along an horizontal line. [sent-68, score-0.366]
29 The goal with this protocol is to bring the cursor to a target that randomly appears either on the left (blue square) or on the right(red square) of the cursor. [sent-69, score-0.359]
30 The target is no further away than 3 positions from the cursor (symbolizing the current position of the robot). [sent-70, score-0.328]
31 This prevents the subject from habituation to one of the stimuli since the cursor reaches the target within a small number of steps. [sent-71, score-0.44]
32 Figure 1 illustrates the protocol with the target (blue) initially positioned 2 steps away on the left side of the cursor (green). [sent-72, score-0.359]
33 An error occurred at step 3) so that the cursor reaches the target in 5 steps. [sent-73, score-0.364]
34 The subjects were asked to imagine a movement of their left hand for the left target and to imagine a movement of their right foot for the right target (note that subject n◦ 1 selected left foot for the left target and right hand for the right target). [sent-75, score-1.083]
35 However, since the subjects had no prior BCI experience, the system was not moving the cursor following the mental commands of the subject, but with an error rate of 20%, to avoid random or totally biased behavior of the cursor. [sent-76, score-0.699]
36 Six healthy volunteer subjects with no prior BCI experience participated in these experiments. [sent-77, score-0.267]
37 After the presentation of the target, the subject focuses on the corresponding mental task until the cursor reached the target. [sent-78, score-0.548]
38 The system moved the cursor with an error rate of 20%; i. [sent-79, score-0.278]
39 , at each step, there was a 20% probability that the cursor moved in the opposite direction. [sent-81, score-0.242]
40 When the cursor reached a target, it briefly turned from green to light green and then a new target was randomly selected by the system. [sent-82, score-0.414]
41 If the cursor didn’t reach the target after 10 steps, a new target was selected. [sent-83, score-0.414]
42 As shown in figure 2, while the subject focuses on a specific mental task, the system delivers a feedback about every 2 seconds. [sent-84, score-0.457]
43 This provides a window just before the feedback for BCI classification and a window just after the feedback for ErrP detection for every single trial. [sent-85, score-0.498]
44 The system delivers a feedback about every 2 seconds, this provides a window just before the feedback for BCI classification and a window just after the feedback for ErrP detection for every single trial. [sent-91, score-0.649]
45 As a new target is presented, the subject focuses on the corresponding mental task until the target is reached. [sent-92, score-0.478]
46 EEG potentials were acquired with a portable system (Biosemi ActiveTwo) by means of a cap with 64 integrated electrodes covering the whole scalp uniformly. [sent-93, score-0.261]
47 Then for off-line mental tasks classification, the power spectrum density (PSD) of EEG channels was estimated over a window of one second just before the feedback. [sent-97, score-0.314]
48 Indeed the actual input vector for the statistical classifier described below is a 150 ms window starting 250 ms after the feedback for channels FCz and Cz. [sent-104, score-0.558]
49 For both mental tasks and ErrP classification, the two different classes (left or right for mental tasks and error or correct for ErrP) are recognized by a Gaussian classifier. [sent-106, score-0.502]
50 Sensory motor rhythm (12-16 Hz) and some beta components are discriminant for all subjects. [sent-116, score-0.273]
51 The most relevant electrodes are in the central area (C3, C4 and Cz) according to the ERD/ERD location for hand and foot movement or imagination. [sent-118, score-0.29]
52 Table 1 shows the classification rates for the two mental tasks and the general BCI accuracy for the 6 subjects and the average of them, it also shows the features (electrodes and frequencies) used for classification. [sent-124, score-0.349]
53 For all 6 subjects, the 12-16 Hz band (sensory motor rhythm (SMR)) appears to be relevant for classification. [sent-125, score-0.226]
54 For subject 2 this peak in the beta band is centered at 20 Hz and for subject 6 it is centered at 30 Hz. [sent-127, score-0.426]
55 Finally subject 4 shows no particular discriminant power in the beta band. [sent-128, score-0.202]
56 During periods of inactivity, brain areas are in a kind of idling state with large populations of neurons firing in synchrony resulting in an increase of amplitude of specific alpha (8-12 Hz) and beta (12-26 Hz) bands. [sent-132, score-0.235]
57 In our case, the most relevant electrodes for all subjects are in the C3, C4 or Cz area. [sent-135, score-0.259]
58 These locations confirm previous studies since C3 and C4 areas usually show ERD/ERS during hands movement or imagination whereas foot movement or imagination are focused in the Cz area [12]. [sent-136, score-0.516]
59 Table 1: Percentages (mean and standard deviations) of correctly recognized single trials for the 2 motor imagination tasks for the 6 subjects and the average of them. [sent-137, score-0.581]
60 All subjects show classification rates of about 70-75% for motor imagination and the general BCI accuracy is 73%. [sent-138, score-0.398]
61 2 * Left foot and Right hand All 6 subjects show classification rates of about 70-75% for motor imagination. [sent-182, score-0.39]
62 However, keeping in mind that first all subjects had no prior BCI experience and second that these figures were obtained exclusively in prediction (i. [sent-185, score-0.195]
63 2 Error-related potentials Figure 4 shows the averages of error trials (red curve), of correct trials (green curve) and the difference error-minus-correct (blue curve) for channel FCz for the six subjects (top). [sent-189, score-0.627]
64 A first small positive peak shows up about ∼230 ms after the feedback (t=0). [sent-190, score-0.401]
65 A negative peak clearly appears ∼290 ms after the feedback for 5 subjects. [sent-191, score-0.401]
66 This negative peak is followed by a positive peak ∼350 ms after the feedback. [sent-192, score-0.36]
67 Finally a second broader negative peak occurs about ∼470 ms after the feedback. [sent-193, score-0.25]
68 All six subjects show similar ErrP time courses whose amplitudes slightly differ from one subject to the other. [sent-195, score-0.39]
69 Furthermore, the fronto-central 5 focus at the occurrence of the different peaks tends to confirm the hypothesis that ErrP are generated in a deep brain region called anterior cingulate cortex [8][9] (see also Section 3. [sent-197, score-0.476]
70 Table 2 reports the recognition rates (mean and standard deviations) for the six subjects plus the average of them. [sent-199, score-0.337]
71 These results show that single-trial recognition of erroneous and correct responses are above 75% and 80%, respectively. [sent-200, score-0.219]
72 The benefit of integrating ErrP detection is obvious since it at least doubles the bit rate for five of the six subjects and the average increase is 124%. [sent-203, score-0.42]
73 Figure 4: (Top) Averages of error trials (red curve), of correct trials (green curve) and the difference errorminus-correct (blue curve) for channel FCz for the six subjects. [sent-204, score-0.374]
74 All six subjects show similar ErrP time courses whose amplitudes slightly differ from one subject to the other. [sent-205, score-0.39]
75 (Bottom) Scalp potentials topographies for the average of the six subjects, at the occurrence of the four described peaks. [sent-206, score-0.3]
76 All focuses are located in frontocentral areas, over the anterior cingulate cortex (ACC). [sent-207, score-0.317]
77 Table 2: Percentages (mean and standard deviations) of correctly recognized error trials and correct trials for the six subjects and the average of them. [sent-208, score-0.592]
78 The benefit of integrating ErrP detection is obvious since it at least doubles the bit rate for five of the six subjects. [sent-211, score-0.258]
79 359 Increase [%] 103 82 187 154 141 101 124 Estimation of intracranial activity Estimating the neuronal sources that generate a given potential map at the scalp surface (EEG) requires the solution of the so-called inverse problem. [sent-269, score-0.266]
80 The ultimate goal is to unmix the signals measured at the scalp and to attribute to each brain area its own estimated temporal activity. [sent-274, score-0.199]
81 This software, known for its zero localization error, was used as a localization tool to estimate the focus of intracranial activity at the occurrence of the four ErrP peaks described in Section 3. [sent-276, score-0.378]
82 Figure 5 shows Talairach slices of localized activity for the grand average of the six subjects at the occurrence of the four described peaks and at the occurrence of a late positive component showing up 650 ms after the feedback. [sent-278, score-0.829]
83 As expected, the areas involved in error processing, namely the pre-supplementary motor area (pre-SMA, Brodmann area 6) and the rostral cingulate zone (RCZ, Brodmann areas 24 & 32) are systematically activated [8][9]. [sent-279, score-0.491]
84 For the second positive peak (350 ms) and mainly for the late positive component (650 ms), parietal areas are also activated. [sent-280, score-0.265]
85 These associative areas (somatosensory association cortex, Brodmann areas 5 & 7) could be related to the fact that the subject becomes aware of the error. [sent-281, score-0.236]
86 It has been proposed that the positive peak was associated with conscious error recognition in case of error potentials elicited in reaction task paradigm [13]. [sent-282, score-0.364]
87 In our case, activation of parietal areas after 350 ms after the feedback agrees with this hypothesis. [sent-283, score-0.398]
88 Figure 5: Talairach slices of localized activity for the grand average of the six subjects at the occurrence of the four peaks described in Section 3. [sent-284, score-0.548]
89 2 and at the occurrence of a late positive component showing up 650 ms after the feedback. [sent-285, score-0.281]
90 Supplementary motor cortex and anterior cingulate cortex are systematically activated. [sent-286, score-0.48]
91 Furthermore, for the second positive peak (350 ms) and mainly for the late positive component (650 ms), parietal areas are also activated. [sent-287, score-0.265]
92 In particular, we have confirmed the existence of a new kind of error-related potential elicited in reaction to an erroneous recognition of the subject’s intention. [sent-290, score-0.249]
93 More importantly, we have shown the feasibility of simultaneously and satisfactorily detecting erroneous responses of the interface and classifying motor imagination for device control at the level of single trials. [sent-291, score-0.432]
94 The preliminary results are very promising and confirm that the online detection of errors is a tool of great benefit, especially for subjects with no prior BCI experience or showing low BCI performance. [sent-296, score-0.267]
95 In parallel, we are exploring how to increase the recognition rate of single-trial erroneous and correct responses. [sent-297, score-0.219]
96 7 In this study we have also shown that, as expected, typical cortical areas involved in error processing such as pre-supplementary motor area and anterior cingulate cortex are systematically activated at the occurrence of the different peaks. [sent-298, score-0.677]
97 The software used for the estimation of the intracranial activity (sLORETA) is only a localization tool. [sent-299, score-0.199]
98 Boosting bit rates and error detection for the classification of fast-paced motor commands based on single-trial EEG analysis. [sent-346, score-0.397]
99 Error-related brain potentials are differently related to awareness of response errors: Evidence from an antisaccade task. [sent-424, score-0.197]
100 High-resolution electroencephalogram: Source estimates of laplacian-transformed somatosensory-evoked potentials using realistic subject head model constructed from magnetic resonance imaging. [sent-434, score-0.203]
wordName wordTfidf (topN-words)
[('errp', 0.563), ('bci', 0.284), ('cursor', 0.242), ('subjects', 0.162), ('mental', 0.153), ('feedback', 0.151), ('motor', 0.143), ('ms', 0.14), ('eeg', 0.14), ('erroneous', 0.124), ('cingulate', 0.122), ('six', 0.116), ('subject', 0.112), ('peak', 0.11), ('commands', 0.106), ('electrodes', 0.097), ('occurrence', 0.093), ('anterior', 0.093), ('imagination', 0.093), ('trials', 0.093), ('potentials', 0.091), ('cz', 0.091), ('hz', 0.088), ('target', 0.086), ('foot', 0.085), ('activity', 0.084), ('interaction', 0.081), ('bcis', 0.079), ('intracranial', 0.076), ('sloreta', 0.076), ('movement', 0.075), ('scalp', 0.073), ('detection', 0.072), ('channels', 0.065), ('window', 0.062), ('areas', 0.062), ('classi', 0.061), ('cortex', 0.061), ('muscular', 0.061), ('mill', 0.06), ('brain', 0.06), ('recognition', 0.059), ('recognized', 0.056), ('frequencies', 0.052), ('beta', 0.049), ('late', 0.048), ('mcfarland', 0.048), ('del', 0.047), ('peaks', 0.047), ('awareness', 0.046), ('babiloni', 0.046), ('fcz', 0.046), ('ferrez', 0.046), ('grand', 0.046), ('parietal', 0.045), ('green', 0.043), ('interfaces', 0.043), ('band', 0.043), ('rm', 0.041), ('focuses', 0.041), ('discriminant', 0.041), ('clinical', 0.041), ('bit', 0.04), ('brodmann', 0.04), ('rhythm', 0.04), ('volunteer', 0.04), ('localization', 0.039), ('imagine', 0.038), ('error', 0.036), ('correct', 0.036), ('rejection', 0.036), ('device', 0.036), ('interface', 0.036), ('dp', 0.035), ('cortical', 0.034), ('tasks', 0.034), ('kind', 0.034), ('delivered', 0.034), ('trial', 0.033), ('signals', 0.033), ('inverse', 0.033), ('area', 0.033), ('experience', 0.033), ('robot', 0.033), ('intent', 0.032), ('healthy', 0.032), ('rehabilitation', 0.032), ('elicited', 0.032), ('asked', 0.031), ('left', 0.031), ('cpz', 0.03), ('dipoles', 0.03), ('doubles', 0.03), ('electrode', 0.03), ('electromagnetic', 0.03), ('focussing', 0.03), ('gestures', 0.03), ('idling', 0.03), ('parc', 0.03), ('purposeful', 0.03)]
simIndex simValue paperId paperTitle
same-paper 1 0.99999982 74 nips-2007-EEG-Based Brain-Computer Interaction: Improved Accuracy by Automatic Single-Trial Error Detection
Author: Pierre Ferrez, José Millán
Abstract: Brain-computer interfaces (BCIs), as any other interaction modality based on physiological signals and body channels (e.g., muscular activity, speech and gestures), are prone to errors in the recognition of subject’s intent. An elegant approach to improve the accuracy of BCIs consists in a verification procedure directly based on the presence of error-related potentials (ErrP) in the EEG recorded right after the occurrence of an error. Six healthy volunteer subjects with no prior BCI experience participated in a new human-robot interaction experiment where they were asked to mentally move a cursor towards a target that can be reached within a few steps using motor imagination. This experiment confirms the previously reported presence of a new kind of ErrP. These “Interaction ErrP” exhibit a first sharp negative peak followed by a positive peak and a second broader negative peak (∼290, ∼350 and ∼470 ms after the feedback, respectively). But in order to exploit these ErrP we need to detect them in each single trial using a short window following the feedback associated to the response of the classifier embedded in the BCI. We have achieved an average recognition rate of correct and erroneous single trials of 81.8% and 76.2%, respectively. Furthermore, we have achieved an average recognition rate of the subject’s intent while trying to mentally drive the cursor of 73.1%. These results show that it’s possible to simultaneously extract useful information for mental control to operate a brain-actuated device as well as cognitive states such as error potentials to improve the quality of the braincomputer interaction. Finally, using a well-known inverse model (sLORETA), we show that the main focus of activity at the occurrence of the ErrP are, as expected, in the pre-supplementary motor area and in the anterior cingulate cortex. 1
2 0.28329444 106 nips-2007-Invariant Common Spatial Patterns: Alleviating Nonstationarities in Brain-Computer Interfacing
Author: Benjamin Blankertz, Motoaki Kawanabe, Ryota Tomioka, Friederike Hohlefeld, Klaus-Robert Müller, Vadim V. Nikulin
Abstract: Brain-Computer Interfaces can suffer from a large variance of the subject conditions within and across sessions. For example vigilance fluctuations in the individual, variable task involvement, workload etc. alter the characteristics of EEG signals and thus challenge a stable BCI operation. In the present work we aim to define features based on a variant of the common spatial patterns (CSP) algorithm that are constructed invariant with respect to such nonstationarities. We enforce invariance properties by adding terms to the denominator of a Rayleigh coefficient representation of CSP such as disturbance covariance matrices from fluctuations in visual processing. In this manner physiological prior knowledge can be used to shape the classification engine for BCI. As a proof of concept we present a BCI classifier that is robust to changes in the level of parietal α -activity. In other words, the EEG decoding still works when there are lapses in vigilance.
3 0.21867406 173 nips-2007-Second Order Bilinear Discriminant Analysis for single trial EEG analysis
Author: Christoforos Christoforou, Paul Sajda, Lucas C. Parra
Abstract: Traditional analysis methods for single-trial classification of electroencephalography (EEG) focus on two types of paradigms: phase locked methods, in which the amplitude of the signal is used as the feature for classification, e.g. event related potentials; and second order methods, in which the feature of interest is the power of the signal, e.g. event related (de)synchronization. The procedure for deciding which paradigm to use is ad hoc and is typically driven by knowledge of the underlying neurophysiology. Here we propose a principled method, based on a bilinear model, in which the algorithm simultaneously learns the best first and second order spatial and temporal features for classification of EEG. The method is demonstrated on simulated data as well as on EEG taken from a benchmark data used to test classification algorithms for brain computer interfaces. 1 1.1
4 0.097794712 37 nips-2007-Blind channel identification for speech dereverberation using l1-norm sparse learning
Author: Yuanqing Lin, Jingdong Chen, Youngmoo Kim, Daniel D. Lee
Abstract: Speech dereverberation remains an open problem after more than three decades of research. The most challenging step in speech dereverberation is blind channel identification (BCI). Although many BCI approaches have been developed, their performance is still far from satisfactory for practical applications. The main difficulty in BCI lies in finding an appropriate acoustic model, which not only can effectively resolve solution degeneracies due to the lack of knowledge of the source, but also robustly models real acoustic environments. This paper proposes a sparse acoustic room impulse response (RIR) model for BCI, that is, an acoustic RIR can be modeled by a sparse FIR filter. Under this model, we show how to formulate the BCI of a single-input multiple-output (SIMO) system into a l1 norm regularized least squares (LS) problem, which is convex and can be solved efficiently with guaranteed global convergence. The sparseness of solutions is controlled by l1 -norm regularization parameters. We propose a sparse learning scheme that infers the optimal l1 -norm regularization parameters directly from microphone observations under a Bayesian framework. Our results show that the proposed approach is effective and robust, and it yields source estimates in real acoustic environments with high fidelity to anechoic chamber measurements.
5 0.082941987 125 nips-2007-Markov Chain Monte Carlo with People
Author: Adam Sanborn, Thomas L. Griffiths
Abstract: Many formal models of cognition implicitly use subjective probability distributions to capture the assumptions of human learners. Most applications of these models determine these distributions indirectly. We propose a method for directly determining the assumptions of human learners by sampling from subjective probability distributions. Using a correspondence between a model of human choice and Markov chain Monte Carlo (MCMC), we describe a method for sampling from the distributions over objects that people associate with different categories. In our task, subjects choose whether to accept or reject a proposed change to an object. The task is constructed so that these decisions follow an MCMC acceptance rule, defining a Markov chain for which the stationary distribution is the category distribution. We test this procedure for both artificial categories acquired in the laboratory, and natural categories acquired from experience. 1
6 0.082145005 122 nips-2007-Locality and low-dimensions in the prediction of natural experience from fMRI
7 0.070611678 110 nips-2007-Learning Bounds for Domain Adaptation
8 0.068495154 154 nips-2007-Predicting Brain States from fMRI Data: Incremental Functional Principal Component Regression
9 0.067666441 137 nips-2007-Multiple-Instance Pruning For Learning Efficient Cascade Detectors
10 0.066992946 203 nips-2007-The rat as particle filter
11 0.056441788 155 nips-2007-Predicting human gaze using low-level saliency combined with face detection
12 0.05448202 3 nips-2007-A Bayesian Model of Conditioned Perception
13 0.051427417 103 nips-2007-Inferring Elapsed Time from Stochastic Neural Processes
14 0.048123542 17 nips-2007-A neural network implementing optimal state estimation based on dynamic spike train decoding
15 0.047558192 127 nips-2007-Measuring Neural Synchrony by Message Passing
16 0.045273773 182 nips-2007-Sparse deep belief net model for visual area V2
17 0.044472672 24 nips-2007-An Analysis of Inference with the Universum
18 0.044221446 56 nips-2007-Configuration Estimates Improve Pedestrian Finding
19 0.043737337 114 nips-2007-Learning and using relational theories
20 0.043196782 85 nips-2007-Experience-Guided Search: A Theory of Attentional Control
topicId topicWeight
[(0, -0.145), (1, 0.057), (2, 0.072), (3, 0.005), (4, 0.003), (5, 0.197), (6, 0.086), (7, 0.121), (8, -0.105), (9, -0.142), (10, -0.025), (11, -0.191), (12, -0.062), (13, -0.07), (14, 0.211), (15, -0.054), (16, -0.002), (17, -0.267), (18, -0.094), (19, -0.035), (20, -0.072), (21, -0.124), (22, 0.014), (23, 0.119), (24, 0.032), (25, -0.11), (26, 0.109), (27, -0.12), (28, 0.071), (29, 0.035), (30, -0.033), (31, -0.033), (32, 0.021), (33, -0.066), (34, 0.052), (35, 0.029), (36, 0.009), (37, 0.01), (38, 0.024), (39, -0.049), (40, 0.027), (41, 0.128), (42, 0.024), (43, 0.008), (44, -0.025), (45, -0.01), (46, 0.037), (47, -0.043), (48, 0.003), (49, 0.021)]
simIndex simValue paperId paperTitle
same-paper 1 0.96001285 74 nips-2007-EEG-Based Brain-Computer Interaction: Improved Accuracy by Automatic Single-Trial Error Detection
Author: Pierre Ferrez, José Millán
Abstract: Brain-computer interfaces (BCIs), as any other interaction modality based on physiological signals and body channels (e.g., muscular activity, speech and gestures), are prone to errors in the recognition of subject’s intent. An elegant approach to improve the accuracy of BCIs consists in a verification procedure directly based on the presence of error-related potentials (ErrP) in the EEG recorded right after the occurrence of an error. Six healthy volunteer subjects with no prior BCI experience participated in a new human-robot interaction experiment where they were asked to mentally move a cursor towards a target that can be reached within a few steps using motor imagination. This experiment confirms the previously reported presence of a new kind of ErrP. These “Interaction ErrP” exhibit a first sharp negative peak followed by a positive peak and a second broader negative peak (∼290, ∼350 and ∼470 ms after the feedback, respectively). But in order to exploit these ErrP we need to detect them in each single trial using a short window following the feedback associated to the response of the classifier embedded in the BCI. We have achieved an average recognition rate of correct and erroneous single trials of 81.8% and 76.2%, respectively. Furthermore, we have achieved an average recognition rate of the subject’s intent while trying to mentally drive the cursor of 73.1%. These results show that it’s possible to simultaneously extract useful information for mental control to operate a brain-actuated device as well as cognitive states such as error potentials to improve the quality of the braincomputer interaction. Finally, using a well-known inverse model (sLORETA), we show that the main focus of activity at the occurrence of the ErrP are, as expected, in the pre-supplementary motor area and in the anterior cingulate cortex. 1
2 0.90001011 106 nips-2007-Invariant Common Spatial Patterns: Alleviating Nonstationarities in Brain-Computer Interfacing
Author: Benjamin Blankertz, Motoaki Kawanabe, Ryota Tomioka, Friederike Hohlefeld, Klaus-Robert Müller, Vadim V. Nikulin
Abstract: Brain-Computer Interfaces can suffer from a large variance of the subject conditions within and across sessions. For example vigilance fluctuations in the individual, variable task involvement, workload etc. alter the characteristics of EEG signals and thus challenge a stable BCI operation. In the present work we aim to define features based on a variant of the common spatial patterns (CSP) algorithm that are constructed invariant with respect to such nonstationarities. We enforce invariance properties by adding terms to the denominator of a Rayleigh coefficient representation of CSP such as disturbance covariance matrices from fluctuations in visual processing. In this manner physiological prior knowledge can be used to shape the classification engine for BCI. As a proof of concept we present a BCI classifier that is robust to changes in the level of parietal α -activity. In other words, the EEG decoding still works when there are lapses in vigilance.
3 0.76313329 173 nips-2007-Second Order Bilinear Discriminant Analysis for single trial EEG analysis
Author: Christoforos Christoforou, Paul Sajda, Lucas C. Parra
Abstract: Traditional analysis methods for single-trial classification of electroencephalography (EEG) focus on two types of paradigms: phase locked methods, in which the amplitude of the signal is used as the feature for classification, e.g. event related potentials; and second order methods, in which the feature of interest is the power of the signal, e.g. event related (de)synchronization. The procedure for deciding which paradigm to use is ad hoc and is typically driven by knowledge of the underlying neurophysiology. Here we propose a principled method, based on a bilinear model, in which the algorithm simultaneously learns the best first and second order spatial and temporal features for classification of EEG. The method is demonstrated on simulated data as well as on EEG taken from a benchmark data used to test classification algorithms for brain computer interfaces. 1 1.1
4 0.58801788 37 nips-2007-Blind channel identification for speech dereverberation using l1-norm sparse learning
Author: Yuanqing Lin, Jingdong Chen, Youngmoo Kim, Daniel D. Lee
Abstract: Speech dereverberation remains an open problem after more than three decades of research. The most challenging step in speech dereverberation is blind channel identification (BCI). Although many BCI approaches have been developed, their performance is still far from satisfactory for practical applications. The main difficulty in BCI lies in finding an appropriate acoustic model, which not only can effectively resolve solution degeneracies due to the lack of knowledge of the source, but also robustly models real acoustic environments. This paper proposes a sparse acoustic room impulse response (RIR) model for BCI, that is, an acoustic RIR can be modeled by a sparse FIR filter. Under this model, we show how to formulate the BCI of a single-input multiple-output (SIMO) system into a l1 norm regularized least squares (LS) problem, which is convex and can be solved efficiently with guaranteed global convergence. The sparseness of solutions is controlled by l1 -norm regularization parameters. We propose a sparse learning scheme that infers the optimal l1 -norm regularization parameters directly from microphone observations under a Bayesian framework. Our results show that the proposed approach is effective and robust, and it yields source estimates in real acoustic environments with high fidelity to anechoic chamber measurements.
5 0.41491583 24 nips-2007-An Analysis of Inference with the Universum
Author: Olivier Chapelle, Alekh Agarwal, Fabian H. Sinz, Bernhard Schölkopf
Abstract: We study a pattern classification algorithm which has recently been proposed by Vapnik and coworkers. It builds on a new inductive principle which assumes that in addition to positive and negative data, a third class of data is available, termed the Universum. We assay the behavior of the algorithm by establishing links with Fisher discriminant analysis and oriented PCA, as well as with an SVM in a projected subspace (or, equivalently, with a data-dependent reduced kernel). We also provide experimental results. 1
6 0.30693528 93 nips-2007-GRIFT: A graphical model for inferring visual classification features from human data
7 0.29627687 127 nips-2007-Measuring Neural Synchrony by Message Passing
8 0.27819642 203 nips-2007-The rat as particle filter
9 0.27066278 137 nips-2007-Multiple-Instance Pruning For Learning Efficient Cascade Detectors
10 0.26678228 125 nips-2007-Markov Chain Monte Carlo with People
11 0.26525936 3 nips-2007-A Bayesian Model of Conditioned Perception
12 0.26031715 25 nips-2007-An in-silico Neural Model of Dynamic Routing through Neuronal Coherence
13 0.25448489 154 nips-2007-Predicting Brain States from fMRI Data: Incremental Functional Principal Component Regression
14 0.23865262 124 nips-2007-Managing Power Consumption and Performance of Computing Systems Using Reinforcement Learning
15 0.23540676 122 nips-2007-Locality and low-dimensions in the prediction of natural experience from fMRI
16 0.22647707 149 nips-2007-Optimal ROC Curve for a Combination of Classifiers
17 0.22615041 110 nips-2007-Learning Bounds for Domain Adaptation
18 0.2230399 71 nips-2007-Discriminative Keyword Selection Using Support Vector Machines
19 0.22056581 59 nips-2007-Continuous Time Particle Filtering for fMRI
20 0.21682005 103 nips-2007-Inferring Elapsed Time from Stochastic Neural Processes
topicId topicWeight
[(4, 0.294), (5, 0.024), (13, 0.03), (16, 0.012), (18, 0.04), (19, 0.137), (21, 0.057), (34, 0.025), (35, 0.022), (47, 0.082), (49, 0.015), (83, 0.086), (85, 0.012), (87, 0.017), (90, 0.067)]
simIndex simValue paperId paperTitle
same-paper 1 0.77666152 74 nips-2007-EEG-Based Brain-Computer Interaction: Improved Accuracy by Automatic Single-Trial Error Detection
Author: Pierre Ferrez, José Millán
Abstract: Brain-computer interfaces (BCIs), as any other interaction modality based on physiological signals and body channels (e.g., muscular activity, speech and gestures), are prone to errors in the recognition of subject’s intent. An elegant approach to improve the accuracy of BCIs consists in a verification procedure directly based on the presence of error-related potentials (ErrP) in the EEG recorded right after the occurrence of an error. Six healthy volunteer subjects with no prior BCI experience participated in a new human-robot interaction experiment where they were asked to mentally move a cursor towards a target that can be reached within a few steps using motor imagination. This experiment confirms the previously reported presence of a new kind of ErrP. These “Interaction ErrP” exhibit a first sharp negative peak followed by a positive peak and a second broader negative peak (∼290, ∼350 and ∼470 ms after the feedback, respectively). But in order to exploit these ErrP we need to detect them in each single trial using a short window following the feedback associated to the response of the classifier embedded in the BCI. We have achieved an average recognition rate of correct and erroneous single trials of 81.8% and 76.2%, respectively. Furthermore, we have achieved an average recognition rate of the subject’s intent while trying to mentally drive the cursor of 73.1%. These results show that it’s possible to simultaneously extract useful information for mental control to operate a brain-actuated device as well as cognitive states such as error potentials to improve the quality of the braincomputer interaction. Finally, using a well-known inverse model (sLORETA), we show that the main focus of activity at the occurrence of the ErrP are, as expected, in the pre-supplementary motor area and in the anterior cingulate cortex. 1
2 0.63907337 179 nips-2007-SpAM: Sparse Additive Models
Author: Han Liu, Larry Wasserman, John D. Lafferty, Pradeep K. Ravikumar
Abstract: We present a new class of models for high-dimensional nonparametric regression and classification called sparse additive models (SpAM). Our methods combine ideas from sparse linear modeling and additive nonparametric regression. We derive a method for fitting the models that is effective even when the number of covariates is larger than the sample size. A statistical analysis of the properties of SpAM is given together with empirical results on synthetic and real data, showing that SpAM can be effective in fitting sparse nonparametric models in high dimensional data. 1
3 0.63086629 39 nips-2007-Boosting the Area under the ROC Curve
Author: Phil Long, Rocco Servedio
Abstract: We show that any weak ranker that can achieve an area under the ROC curve slightly better than 1/2 (which can be achieved by random guessing) can be efficiently boosted to achieve an area under the ROC curve arbitrarily close to 1. We further show that this boosting can be performed even in the presence of independent misclassification noise, given access to a noise-tolerant weak ranker.
4 0.5908215 106 nips-2007-Invariant Common Spatial Patterns: Alleviating Nonstationarities in Brain-Computer Interfacing
Author: Benjamin Blankertz, Motoaki Kawanabe, Ryota Tomioka, Friederike Hohlefeld, Klaus-Robert Müller, Vadim V. Nikulin
Abstract: Brain-Computer Interfaces can suffer from a large variance of the subject conditions within and across sessions. For example vigilance fluctuations in the individual, variable task involvement, workload etc. alter the characteristics of EEG signals and thus challenge a stable BCI operation. In the present work we aim to define features based on a variant of the common spatial patterns (CSP) algorithm that are constructed invariant with respect to such nonstationarities. We enforce invariance properties by adding terms to the denominator of a Rayleigh coefficient representation of CSP such as disturbance covariance matrices from fluctuations in visual processing. In this manner physiological prior knowledge can be used to shape the classification engine for BCI. As a proof of concept we present a BCI classifier that is robust to changes in the level of parietal α -activity. In other words, the EEG decoding still works when there are lapses in vigilance.
5 0.56491327 173 nips-2007-Second Order Bilinear Discriminant Analysis for single trial EEG analysis
Author: Christoforos Christoforou, Paul Sajda, Lucas C. Parra
Abstract: Traditional analysis methods for single-trial classification of electroencephalography (EEG) focus on two types of paradigms: phase locked methods, in which the amplitude of the signal is used as the feature for classification, e.g. event related potentials; and second order methods, in which the feature of interest is the power of the signal, e.g. event related (de)synchronization. The procedure for deciding which paradigm to use is ad hoc and is typically driven by knowledge of the underlying neurophysiology. Here we propose a principled method, based on a bilinear model, in which the algorithm simultaneously learns the best first and second order spatial and temporal features for classification of EEG. The method is demonstrated on simulated data as well as on EEG taken from a benchmark data used to test classification algorithms for brain computer interfaces. 1 1.1
7 0.45240942 37 nips-2007-Blind channel identification for speech dereverberation using l1-norm sparse learning
8 0.44834447 24 nips-2007-An Analysis of Inference with the Universum
9 0.44139612 122 nips-2007-Locality and low-dimensions in the prediction of natural experience from fMRI
10 0.44049805 167 nips-2007-Regulator Discovery from Gene Expression Time Series of Malaria Parasites: a Hierachical Approach
11 0.43684706 53 nips-2007-Compressed Regression
12 0.43429857 202 nips-2007-The discriminant center-surround hypothesis for bottom-up saliency
13 0.43388888 154 nips-2007-Predicting Brain States from fMRI Data: Incremental Functional Principal Component Regression
14 0.43101332 93 nips-2007-GRIFT: A graphical model for inferring visual classification features from human data
15 0.42768228 161 nips-2007-Random Projections for Manifold Learning
16 0.42398259 18 nips-2007-A probabilistic model for generating realistic lip movements from speech
17 0.42267779 177 nips-2007-Simplified Rules and Theoretical Analysis for Information Bottleneck Optimization and PCA with Spiking Neurons
18 0.4224042 45 nips-2007-Classification via Minimum Incremental Coding Length (MICL)
19 0.42033261 34 nips-2007-Bayesian Policy Learning with Trans-Dimensional MCMC
20 0.42006558 138 nips-2007-Near-Maximum Entropy Models for Binary Neural Representations of Natural Images