Alon Fishbach

Action selection and refinement in subcortical loops through basal ganglia and cerebellum

Subcortical loops through the basal ganglia and the cerebellum form computationally powerful distributed processing modules (DPMs). This paper relates the computational features of a DPMs loop through the basal ganglia to experimental results for two kinds of natural action selection. First, functional imaging during a serial order recall task was used to study human brain activity during the selection of sequential actions from working memory. Second, microelectrode recordings from monkeys trained in a step-tracking task were used to study the natural selection of corrective submovements. Our DPM-based model assisted in the interpretation of puzzling data from both of these experiments. We come to posit that the many loops through the basal ganglia each regulate the embodiment of pattern formation in a given area of cerebral cortex. This operation serves to instantiate different kinds of action (or thought) mediated by different areas of cerebral cortex. We then use our findings to formulate a model of the aetiology of schizophrenia.

Primary auditory cortex of cats: feature detection or something else?

Abstract.Neurons in sensory cortices are often assumed to be “feature detectors”, computing simple and then successively more complex features out of the incoming sensory stream. These features are somehow integrated into percepts. Despite many years of research, a convincing candidate for such a feature in primary auditory cortex has not been found. We argue that feature detection is actually a secondary issue in understanding the role of primary auditory cortex. Instead, the major contribution of primary auditory cortex to auditory perception is in processing previously derived features on a number of different timescales. We hypothesize that, as a result, neurons in primary auditory cortex represent sounds in terms of auditory objects rather than in terms of feature maps. According to this hypothesis, primary auditory cortex has a pivotal role in the auditory system in that it generates the representation of auditory objects to which higher auditory centers assign properties such as spatial location, source identity, and meaning.

Deciding when and how to correct a movement: discrete submovements as a decision making process.

Rapid reaching movements of human and non-human primates are often characterized by irregular multi-peaked velocity profiles. How to interpret these irregularities is still under debate. While some reports assert that these irregularities are the result of a continuous controller interacting with the environment, we and others hold that the velocity irregularities are evidence for a controller that produces discrete movement corrections. Here we analyze rapid pronation/supination wrist movements in monkey during a 1D step-tracking task, where visual perturbations of the target were randomly introduced at movement onset. We use our recently introduced algorithm (Fishbach et al. in Exp Brain Res 164:442–457, 2005) to decompose an irregular movement into a primary movement and one or more discrete, corrective submovements. We first show that the visual perturbation has almost no effect on primary movements. In contrast, this perturbation influences the type and the extent of the corrective submovements that often follow primary movements. Secondly, we show that the highly variable timing of overlapping submovements does not depend directly on the visual perturbation but rather on an estimate of the movement error and on the movement’s extent-to-go at the time of correction initiation. These results are consistent with a forward-model based intermittent controller with a non-linearity that depends both on a prediction of the magnitude and direction of the movement’s error and on its variance. Corrections are initiated only when the predicted error is statistically significant. A simple abstract model that implements these principles accounts for the type and timing of the corrections observed in our data.

Kinematic properties of on-line error corrections in the monkey

Despite the abundant experimental evidence for the irregular, multipeaked velocity profiles that often characterize rapid human limb movements, there is currently little agreement on how to interpret these phenomena. While in some studies these irregularities have been interpreted as reflecting a continuous control process, in others the irregularities are considered to be evidence for the existence of discrete movement primitives that are initiated by an intermittent controller. Here we introduce a novel “soft symmetry” method for analyzing irregular movements and decomposing them into their discrete movement primitives. We applied this method to analyze rapid pronation/supination wrist movements in monkeys during a one-dimensional tracking task. We showed that the properties of the extracted overlapping submovements (OSMs) were very similar to those of single, regular movements, despite the fact that the decomposition algorithm did not restrict the extracted submovements to a particular shape. In addition we showed that the movement primitives corrected preceding primitives and that the correction initiation time was highly variable, and thus could not be explained by the relatively fixed sensorimotor delay. These results argue against the interpretation of movement irregularities as reflecting a continuous control process and reinforce the hypothesis that movement irregularities result from an intermittent control mechanism. Demonstrating these phenomena in non-human primates will allow neurophysiological investigation of the neural mechanisms involved in these corrections.

Functional reorganization of upper-body movement after spinal cord injury

Survivors of spinal cord injury need to reorganize their residual body movements for interacting with assistive devices and performing activities that used to be easy and natural. To investigate movement reorganization, we asked subjects with high-level spinal cord injury (SCI) and unimpaired subjects to control a cursor on a screen by performing upper-body motions. While this task would be normally accomplished by operating a computer mouse, here shoulder motions were mapped into the cursor position. Both the control and the SCI subjects were rapidly able to reorganize their movements and to successfully control the cursor. The majority of the subjects in both groups were successful in reducing the movements that were not effective at producing cursor motions. This is inconsistent with the hypothesis that the control system is merely concerned with the accurate acquisition of the targets and is unconcerned with motions that are not relevant to this goal. In contrast, our findings suggest that subjects can learn to reorganize coordination so as to increase the correspondence between the subspace of their upper-body motions with the plane in which the controlled cursor moves. This is effectively equivalent to constructing an inverse internal model of the map from body motions to cursor motions, established by the experiment. These results are relevant to the development of interfaces for assistive devices that optimize the use of residual voluntary control and enhance the learning process in disabled users, searching for an easily learnable map between their body motor space and control space of the device.

Learning Algorithms for Human–Machine Interfaces

The goal of this study is to create and examine machine learning algorithms that adapt in a controlled and cadenced way to foster a harmonious learning environment between the user and the controlled device. To evaluate these algorithms, we have developed a simple experimental framework. Subjects wear an instrumented data glove that records finger motions. The high-dimensional glove signals remotely control the joint angles of a simulated planar two-link arm on a computer screen, which is used to acquire targets. A machine learning algorithm was applied to adaptively change the transformation between finger motion and the simulated robot arm. This algorithm was either LMS gradient descent or the Moore-Penrose (MP) pseudoinverse transformation. Both algorithms modified the glove-to-joint angle map so as to reduce the endpoint errors measured in past performance. The MP group performed worse than the control group (subjects not exposed to any machine learning), while the LMS group outperformed the control subjects. However, the LMS subjects failed to achieve better generalization than the control subjects, and after extensive training converged to the same level of performance as the control subjects. These results highlight the limitations of coadaptive learning using only endpoint error reduction.

Body machine interface: Remapping motor skills after spinal cord injury

The goal of a body-machine interface (BMI) is to map the residual motor skills of the users into efficient patterns of control. The interface is subject to two processes of learning: while users practice controlling the assistive device, the interface modifies itself based on the users residual abilities and preferences. In this study, we combined virtual reality and movement capture technologies to investigate the reorganization of movements that occurs when individuals with spinal cord injury (SCI) are allowed to use a broad spectrum of body motions to perform different tasks. Subjects, over multiple sessions, used their upper body movements to engage in exercises that required different operational functions such as controlling a keyboard for playing a videogame, driving a simulated wheelchair in a virtual reality (VR) environment, and piloting a cursor on a screen for reaching targets. In particular, we investigated the possibility of reducing the dimensionality of the control signals by finding repeatable and stable correlations of movement signals, established both by the presence of biomechanical constraints and by learned patterns of coordination. The outcomes of these investigations will provide guidance for further studies of efficient remapping of motor coordination for the control of assistive devices and are a basis for a new training paradigm in which the burden of learning is significantly removed from the impaired subjects and shifted to the devices.

Seeing versus Believing: Conflicting Immediate and Predicted Feedback Lead to Suboptimal Motor Performance

Reaching hand movements tend to follow straight paths. Previous work has suggested that when visual feedback is perturbed such that straight hand motions are seen as curved motions, the motor system adapts to restore straight visual motion. We show that under a nonlinear visuomotor transformation, one that maps straight hand motions to high-curvature motions of a visual cursor, reaching movements do not converge with practice toward a straight path of either the hand or the cursor. Instead, hand trajectories converged to a repeatable and characteristic curved shape. We propose a new computational model in which the adapted trajectories are obtained by minimizing a cost function composed of two terms. The first term enforces hand-movement smoothness. The second term penalizes average visual aiming error, which is the instantaneous discrepancy between the direction of the hand movement and the direction of the vector that points from the cursor to the target. Our results are consistent with the models predictions and demonstrate a persistent effect of the predicted feedback of direction errors despite the possibility of producing smoother hand motions by ignoring it.

Functional reorganization of upper-body movements for wheelchair control

In general, survivors of neuromotor disorders and injuries need to reorganize their body movements in order to achieve goals that used to be easy and natural. Often, disabled people are offered the option to control assistive devices that will facilitate the recovery of independence and capability in their daily lives. The knowledge acquired during the last few years in the motor control field can be used to study and enhance this learning process. Furthermore, this knowledge may aid in finding methods for optimizing the use of residual voluntary muscular control in disabled users and searching for an easily learnable map between body motor space and devices control space. To investigate movement reorganization we asked healthy subjects to control a cursor performing a reaching task using shoulders and upper arm movements. These movements were mapped to a lower dimensional space by principal components analysis and were used to control the cursor. We found that all subjects were able to learn to control the cursor with ease and precision while reducing the proportion of ineffective body movement components in favor of the components that mapped directly into the control space. Moreover, with practice the movements of the controlled device – the cursor - became faster, smother, more precise and with a nearly symmetric speed profile.

Spectral sound gap filling

We present a new method for automatically filling in gaps of textural sounds. Our approach is to transform the signal to the time-frequency space, fill in the gap, and apply the inverse transform to reconstruct the result. The complex spectrogram of the signal is partitioned into separate overlapping frequency bands. Each band is fragmented by segmentation of the time-frequency space and a partition of the spectrogram in time, and filled in with complex fragments by example. We demonstrate our method by filling in gaps of various types of textural sounds.