Perceptual scaling of voice identity: common dimensions for different vowels and speakers

Oliver Baumann, Pascal Belin* ,
Queensland Brain Institute, The University of Queensland
* University of Glasgow

ABSTRACT

The aims of our study were: (1) to determine if the acoustical parameters used by normal subjects to discriminate between different speakers vary when comparisons are made between pairs of two of the same or different vowels, and if they are different for male and female voices; (2) to ask whether individual voices can reasonably be represented as points in a low-dimensional perceptual space such that similarly sounding voices are located close to one another. Subjects were presented with pairs of voices from 16 male and 16 female speakers uttering the three French vowels ‘‘a’’, ‘‘i’’ and ‘‘u’’ and asked to give speaker similarity judgments. Multidimensional analyses of the similarity matrices were performed separately for male and female voices and for three types of comparisons: same vowels, different vowels and overall average. The resulting dimensions were then interpreted a posteriori in terms of relevant acoustical measures. For both male and female voices, a two-dimensional perceptual space was found to be most appropriate, with axes largely corresponding to contributions of the larynx (pitch) and supra-laryngeal vocal tract (formants), mirroring the two largely independent components of source and filter in voice production. These perceptual spaces of male and female voices and their corresponding voice samples are available at: http://vnl.psy.gla.ac.uk section Resources.

Reference

Baumann O. & Belin P. Perceptual scaling of voice identity: common dimensions for different vowels and speakers. Psychological Research 74: 110-120 (2010)

PDF (200K)

Fig 1. The two-dimensional voice space: a spatial model derived with the ALSCAL procedure from dissimilarity ratings on 16 male voices by 10 subjects (averaged over all types of comparisons and vowels).

Fig 2. The two-dimensional voice space: a spatial model derived with the ALSCAL procedure from dissimilarity ratings on 16 female voices by 10 subjects (averaged over all types of comparisons and vowels). 

Scaling of Neural Responses to Visual and Auditory Motion in the Human Cerebellum

Oliver Baumann and Jason B. Mattingley,
Queensland Brain Institute and School of Psychology,
The University of Queensland

ABSTRACT

The human cerebellum contains approximately half of all the neurons within the cerebrum, yet most experimental work in human neuroscience over the last century has focused exclusively on the structure and functions of the forebrain. The cerebellum has an undisputed role in a range of motor functions (Thach et al., 1992), but its potential contributions to sensory and cognitive processes are widely debated (Stoodley and Schmahmann, 2009). Here we used functional magnetic resonance imaging to test the hypothesis that the human cerebellum is involved in the acquisition of auditory and visual sensory data. We monitored neural activity within the cerebellum while participants engaged in a task that required them to discriminate the direction of a visual or auditory motion signal in noise. We identified a distinct set of cerebellar regions that were differentially activated for visual stimuli (vermal lobule VI and right-hemispheric lobule X) and auditory stimuli (right-hemispheric lobules VIIIA and VIIIB and hemispheric lobule VI bilaterally). In addition, we identified a region in left crus I in which activity correlated significantly with increases in the perceptual demands of the task (i.e., with decreasing signal strength), for both auditory and visual stimuli. Our results support suggestions of a role for the cerebellum in the processing of auditory and visual motion and suggest that parts of cerebellar cortex are concerned with tracking movements of objects around the animal, rather than with controlling movements of the animal itself (Paulin, 1993).

Reference

Baumann O. & Mattingley J.B. Scaling of neural responses to visual and auditory motion in the human cerebellum. The Journal of Neuroscience 30: 4489-4495 (2010a)

PDF (466K)

Fig 1. Human cerebellar anatomy shown in sagittal and coronal planes. The locations of anatomical regions (nomenclature according to Schmahmann et al., 2000) were derived using the probabilistic atlas of the cerebellum by Diedrichsen et al. (2009).

Dissociable neural circuits for encoding and retrieval of object locations during active navigation in humans

Oliver Baumann, Edgar Chan, Jason B. Mattingley,
The University of Queensland

ABSTRACT

Several cortical and subcortical circuits have been implicated in object location memory and navigation. Uncertainty remains, however, about which neural circuits are involved in the distinct processes of encoding and retrieval during active navigation through three-dimensional space. We used functional magnetic resonance imaging (fMRI) to measure neural responses as participants learned the location of a single target object relative to a small set of landmarks. Following a delay, the target was removed and participants were required to navigate back to its original position. The relative and absolute locations of landmarks and the target object were changed on every trial, so that participants had to learn a novel arrangement for each spatial scene. At encoding, greater activity within the right hippocampus and the parahippocampal gyrus bilaterally predicted more accurate navigation to the hidden target object in the retrieval phase. By contrast, during the retrieval phase, more accurate performance was associated with increased activity in the left hippocampus and the striatum bilaterally. Dividing participants into good and poor navigators, based upon behavioural performance, revealed greater striatal activity in good navigators during retrieval, perhaps reflecting superior procedural learning in these individuals. By contrast, the poor navigators showed stronger left hippocampal activity, suggesting reliance on a less effective verbal or symbolic code by this group. Our findings suggest separate neural substrates for the encoding and retrieval stages of object location memory during active navigation, which are further modulated by participants' overall navigational ability.

Reference

O Baumann, E Chan, J B Mattingley (2010). Dissociable neural circuits for encoding and retrieval of object locations during active navigation in humans NeuroImage 49 2816–2825

PDF (1,035K)

Fig 1. Schematic of the virtual environment used in the navigation task. (a) Example display of the virtual environment. (b) Sequence of events in a typical experimental trial.

Fig 2.  Three-dimensional rendered MR images showing mean BOLD activity from the whole-brain analysis of experimental minus baseline conditions. Red represents activity during the encoding phase; green represents activity during the retrieval phase; yellow indicates overlaps.

The Information Content of Panoramic Images I: The Rotational Errors and the Similarity of Views in Rectangular Experimental Arenas

Wolfgang Stürzl, Allen Cheung, and Jochen Zeil,
The Australian National University;
Ken Cheng, Macquarie University

ABSTRACT

Animals relocating a target corner in a rectangular space often make rotational errors searching not only at the target corner but also at the diagonally opposite corner. The authors tested whether view-based navigation can explain rotational errors by recording panoramic snapshots at regularly spaced locations in a rectangular box. The authors calculated the global image difference between the image at each location and the image recorded at a target location in 1 of the corners, thus creating a 2-dimensional map of image differences. The authors found the most pronounced minima of image differences at the target corner and the diagonally opposite corner— conditions favouring rotational errors. The authors confirmed these results in virtual reality simulations and showed that the relative salience of different visual cues determines whether image differences are dominated by geometry or by features. The geometry of space is thus implicitly contained in panoramic images and does not require explicit computation by a dedicated module. A testable prediction is that animals making rotational errors in rectangular spaces are guided by remembered views.

Reference

W Stürzl, A Cheung, J Zeil, K Cheng (2008). The Information Content of Panoramic Images I: The Rotational Errors and the Similarity of Views in Rectangular Experimental Arenas. Journal of Experimental Psychology: Animal Behavior Processes 2008, Vol. 34, No. 1, 1–14

PDF (2,596K)

Fig 1. The image difference function (IDF) in a box with four black walls without (on the left) and with distinct corner features (on the right). (A) Panoramic views of the box. The target location is marked with a black circle and the condition of the box walls is shown in the color of the frame. (B and C). (D) Difference functions over the horizontal and vertical transects. (E) Local image differences are determined only for the 8 neighboring locations. (F) Local image differences are determined for 24 neighboring locations.

The Information Content of Panoramic Images II: View-Based Navigation in Nonrectangular Experimental Arenas

Allen Cheung, Wolfgang Stürzl, and Jochen Zeil,
The Australian National University;
Ken Cheng, Macquarie University.

ABSTRACT

Two recent studies testing navigation of rats in swimming pools have posed problems for any account of the use of purely geometric properties of space in navigation (M. Graham, M. A. Good, A. McGregor, & J. M. Pearce, 2006; J. M. Pearce, M. A. Good, P. M. Jones, & A. McGregor, 2004). The authors simulated 1 experiment from each study in a virtual reality environment to test whether experimental results could be explained by view-based navigation. The authors recorded a reference image at the target location and then determined global panoramic image differences between this image and images taken at regularly spaced locations throughout the arena. A formal model, in which an agent attempts to minimize image differences between the reference image and current views, generated trajectories that could be compared with the search performance of rats. For both experiments, this model mimics many aspects of rat behavior. View-based navigation provides a sufficient and parsimonious explanation for a range of navigational behaviors of rats under these experimental conditions.

Reference

A Cheung, W Stürzl, J Zeil, K Cheng.(2008) The Information Content of Panoramic Images II: View-Based Navigation in Nonrectangular Experimental Arenas. Journal of Experimental Psychology: Animal Behavior Processes 2008, Vol. 34, No. 1, 15–30

PDF (1,362K)

Fig 1. Rats were trained to find a hidden platform close to a corner of a rectangular swimming pool (A). The rats were then tested in a kite-shaped pool (B). (C)The three-dimensional representation of the image difference function (IDF). (D)The two-dimensional representation. (E) Local transects through the local minima of the IDF. (F) The local slopes of the IDF.

Fig 2. Comparison of rat first-choice behavior with simulation results.

From Behaviour to Brain Dynamics

Allen Cheung, Queensland Brain Institute

ABSTRACT

It is well accepted that medium to long range navigation requires the use of an external directional reference i.e. a compass. Cheung et al (2007) recently demonstrated through theory and simulation the quantitative significance of the compass. It was shown that navigating agents using and not using a compass could be differentiated on the basis of the population behaviour. In the current work, theory and simulation results will be presented on ways to characterize individual paths on the basis of whether the system was using an external directional reference. Thus it is demonstrated that important information concerning the neural input used by a navigating animal may be inferred probabilistically from its behaviour.

Reference

M Marinaro, S Scarpetta, and Y Yamaguchi (Eds.): Dynamic Brain, LNCS 5286, pp. 91–95, 2008. łc Springer-Verlag Berlin Heidelberg 2008

PDF (341K)

Fig 1. (A) Graphical example of an unknown mixture of simulated idiothetic and allothetic directed walks. (B) Set of 30 paths of 20 steps, the decision function (Eqn 6) made the correct decision in 28 out of 30 paths.

Fig 2.  Test performance of the decision function using simulated IDWs and ADWs.

Animal Navigation: General Properties of Directed Walks

Allen Cheung and Mandyam V. Srinivasan,
The University of Queensland;
Shaowu Zhang, University of Canberra;
Christian Stricker, Australian National University.

ABSTRACT

The ability to locomote is a defining characteristic of all animals. Yet, all but the most trivial forms of navigation are poorly understood. Here we report and discuss the analytical results of an in-depth study of a simple navigation problem. In principle, there are two strategies for navigating a straight course. One is to use an external directional reference and to continually reorient with reference to it. The other is to monitor body rotations from internal sensory information only. We showed previously that, at least for simple representations of locomotion, the first strategy will enable an animal or mobile agent to move arbitrarily far away from its starting point, but the second strategy will not do so, even after an infinite number of steps. This paper extends and generalizes the earlier results by demonstrating that these findings are true even when a very general model of locomotion is used. In this general model, error components within individual steps are not independent, and directional errors may be biased. In the absence of a compass, the expected path of a directed walk in general approximates a logarithmic spiral. Some examples are given to illustrate potential applications of the quantitative results derived here. Motivated by the analytical results developed in this work, a nomenclature for directed walks is proposed and discussed. Issues related to path integration in mammals and robots, and measuring the curvature of a noisy path are also addressed using directed walk theory.

Published Online: 10 September 2008

Reference

A Cheung, M V Srinivasan, S Zhang, C Stricker (2008) Animal Navigation: General Properties of Directed Walks. Biological Cybernetics (2008) 99:197–217 DOI 10.1007/s00422-008-0251-z

PDF (610K)

Fig 1. Diagrammatic representation of a general elementary step during an allothetic directed walk (a), and an idiothetic directed walk (b).

Fig 2. Comparison of probability density functions following a 500 step directed walk in the presence and absence of a compass. (b) Using a compass. (c) Without a compass.

Obstacle Avoidance in Cluttered Environments using Optic Flow

Matthew A. Garratt, University of New South Wales; Allen Cheung, The University of Queensland

ABSTRACT

Autonomous navigation of a flying vehicle in a cluttered environment has been simulated to demonstrate novel yet simple techniques for avoiding obstacles using optic flow and image loom signals whilst successfully navigating a vehicle towards a goal. The simulation makes use of simulated flow fields calculated on cameras looking downwards, forwards, rearwards and sidewards. This computation could be carried out using a single processor networked with multiple remote cameras or using separate optic flow sensors oriented in each direction.

Reference

M Garratt, A Cheung (2009) Obstacle Avoidance in Cluttered Environments using Optic Flow. Australasian Conference on Robotics and Automation (ACRA), December 2-4, 2009, Sydney.

PDF (1,719K)

Fig 1. Trajectory of helicopter in horizontal plane within an 8m×16m arena for (a) heading aligned with longer walls and (b) offset by 22.5˚ from the longer walls.

Which coordinate system for modelling path integration?

Robert J. Vickerstaff, University of Canterbury; Allen Cheung, The University of Queensland

ABSTRACT

Path integration is a navigation strategy widely observed in nature where an animal maintains a running estimate, called the home vector, of its location during an excursion. Evidence suggests it is both ancient and ubiquitous in nature, and has been studied for over a century. In that time, canonical and neural network models have flourished, based on a wide range of assumptions, justifications and supporting data. Despite the importance of the phenomenon, consensus and unifying principles appear lacking. A fundamental issue is the neural representation of space needed for biological path integration. This paper presents a scheme to classify path integration systems on the basis of the way the home vector records and updates the spatial relationship between the animal and its home location. Four extended classes of coordinate systems are used to unify and review both canonical and neural network models of path integration, from the arthropod and mammalian literature. This scheme demonstrates analytical equivalence between models which may otherwise appear unrelated, and distinguishes between models which may superficially appear similar. A thorough analysis is carried out of the equational forms of important facets of path integration including updating, steering, searching and systematic errors, using each of the four coordinate systems. The type of available directional cue, namely allothetic or idiothetic, is also considered. It is shown that on balance, the class of home vectors which includes the geocentric Cartesian coordinate system, appears to be the most robust for biological systems. A key conclusion is that deducing computational structure from behavioural data alone will be difficult or impossible, at least in the absence of an analysis of random errors. Consequently it is likely that further theoretical insights into path integration will require an in- depth study of the effect of noise on the four classes of home vectors.

Reference

R J Vickerstaff, A Cheung (2009). Which coordinate system for modelling path integration? Journal of Theoretical Biology 263 (2010) 242–261

PDF (619K)

Fig 1. Four ways to represent the same spatial relationship between animal and home. ‘A’ is the animal’s location. ‘H’ shows the home location. Shown are HVs for each of the four ‘standard’ coordinate systems considered in this paper:(a) geocentric Cartesian (GC), (b) geocentric polar (GP), (c) egocentric Cartesian(EC), (d) egocentric polar (EP).

Fig 2. Homing trajectory generated by the searching model after an L-shaped outwards excursion.

Fig 3. Search density profile of a single search pattern lasting for 10⁶s.

Hybrid robot control and SLAM for persistent navigation and mapping

Michael Milford (a, b, c) and Gordon Wyeth (a), 
(a) School of Engineering Systems, Queensland University of Technology,
(b) Queensland Brain Institute, The University of Queensland,
(c) School of Information Technology and Electrical Engineering, The University of Queensland

ABSTRACT

For a mobile robot to operate autonomously in real-world environments, it must have an effective control system and a navigation system capable of providing robust localization, path planning and path execution. In this paper we describe work investigating synergies between mapping and control systems. We have integrated development of a control system for navigating mobile robots and a robot SLAM system. The control system is hybrid in nature and tightly coupled with the SLAM system; it uses a combination of high and low level deliberative and reactive control processes to perform obstacle avoidance, exploration, global navigation and recharging, and draws upon the map learning and localization capabilities of the SLAM system. The effectiveness of this hybrid, multi-level approach was evaluated in the context of a delivery robot scenario. Over a period of two weeks the robot performed 1143 delivery tasks to 11 different locations with only one delivery failure (from which it recovered), travelled a total distance of more than 40 km, and recharged autonomously a total of 23 times. In this paper we describe the combined control and SLAM system and discuss insights gained from its successful application in a real-world context.

Reference

Robotics and Autonomous Systems Volume 58, Issue 9, 30 September 2010, Pages 1096-1104 Hybrid Control for Autonomous Systems

PDF (1780K)

Learning Spatial concepts from RatSLAM Representations

Michael Milford, Ruth Schulz, David Prasser, Gordon Wyeth and Janet Wiles. The University of Queensland.

ABSTRACT

RatSLAM is a biologically-inspired visual SLAM and navigation system that has been shown to be effective indoors and outdoors on real robots. The spatial representation at the core of RatSLAM, the experience map, forms in a distributed fashion as the robot learns the environment. The activity in RatSLAM’s experience map possesses some geometric properties, but still does not represent the world in a human readable form. A new system, dubbed RatChat, has been introduced to enable meaningful communication with the robot. The intention is to use the “language games” paradigm to build spatial concepts that can be used as the basis for communication. This paper describes the first step in the language game experiments, showing the potential for meaningful categorization of the spatial representations in RatSLAM.

Reference

Robotics and Autonomous Systems Volume 55, Issue 5, 31 May 2007, Pages 403-410 From Sensors to Human Spatial Concepts

PDF (927K)

Fig 1. Floor plan of the area used for the experiment and the approximate trajectory of the robot.

Fig 2. Trajectory of the most highly activated pose cell during the experiment.

Fig 3. The experience map. The map is continuous and has a high degree of correspondence to the spatial arrangement of the environment shown in Fig 1.

Persistent Navigation and Mapping using a Biologically Inspired SLAM System

Michael Milford and Gordon Wyeth, The University of Queensland

ABSTRACT

The challenge of persistent navigation and mapping is to develop an autonomous robot system that can simultaneously localize, map and navigate over the lifetime of the robot with little or no human intervention. Most solutions to the simultaneous localization and mapping (SLAM) problem aim to produce highly accurate maps of areas that are assumed to be static. In contrast, solutions for persistent navigation and mapping must produce reliable goal-directed navigation outcomes in an environment that is assumed to be in constant flux. We investigate the persistent navigation and mapping problem in the context of an autonomous robot that performs mock deliveries in a working office environment over a two-week period. The solution was based on the biologically inspired visual SLAM system, RatSLAM. RatSLAM performed SLAM continuously while interacting with global and local navigation systems, and a task selection module that selected between exploration, delivery, and recharging modes. The robot performed 1,143 delivery tasks to 11 different locations with only one delivery failure (from which it recovered), travelled a total distance of more than 40 km over 37 hours of active operation, and recharged autonomously a total of 23 times.

Published Online: 21 July, 2009 as doi:10.1177/0278364909340592

Reference

M Milford, G Wyeth (2009). Persistent Navigation and Mapping using a Biologically Inspired SLAM System. The International Journal of Robotics Research OnlineFirst.

PDF (2,194K)

Fig 1. Experience map pruning. Experiences are removed to maintain a one experience per grid square density.

Fig 2. Vision hardware.

Fig 3. (a) A photo of the robot in the environment during the experiments and (c)–(d) a schematic showing the local navigation process.

Mapping a Suburb With a Single Camera Using a Biologically Inspired SLAM System

Michael Milford and Gordon Wyeth, The University of Queensland

ABSTRACT

This paper describes a biologically inspired approach to vision-only simultaneous localization and mapping (SLAM) on ground-based platforms. The core SLAM system, dubbed RatSLAM, is based on computational models of the rodent hippocampus, and is coupled with a lightweight vision system that provides odometry and appearance information. RatSLAM builds a map in an online manner, driving loop closure and relocalization through sequences of familiar visual scenes. Visual ambiguity is managed by maintaining multiple competing vehicle pose estimates, while cumulative errors in odometry are corrected after loop closure by a map correction algorithm. We demonstrate the mapping performance of the system on a 66 km car journey through a complex suburban road network. Using only a web camera operating at 10 Hz, RatSLAM generates a coherent map of the entire environment at real-time speed, correctly closing more than 51 loops of up to 5 km in length. structure in neuron ensemble activity.

Reference

M Milford, G Wyeth (2008). Mapping a Suburb With a Single Camera Using a Biologically Inspired SLAM System. IEEE TRANSACTIONS ON ROBOTICS, VOL. 24, NO. 5.

PDF (1902K)

Fig 1. (a) Excitatory (arrows), inhibitory (round), and self-motion connections for a continuous attractor network representation of head direction cells. (b) A stable activity packet centered at 120◦.

Fig 3. Broad connectivity between functional regions in the RatSLAM system.

Single Camera Vision-Only SLAM on a Suburban Road Network  

Michael Milford and Gordon Wyeth,
The University of Queensland

ABSTRACT

Simultaneous Localization And Mapping (SLAM) is one of the major challenges in mobile robotics. Probabilistic techniques using high-end range finding devices are well established in the field, but recent work has investigated vision only approaches. This paper presents a method for generating approximate rotational and translation velocity information from a single vehicle-mounted consumer camera, without the computationally expensive process of tracking landmarks. The method is tested by employing it to provide the odometric and visual information for the RatSLAM system while mapping a complex suburban road network. RatSLAM generates a coherent map of the environment during an 18 km long trip through suburban traffic at speeds of up to 60 km/hr. This result demonstrates the potential of ground-based vision-only SLAM using low cost sensing and computational hardware.

Reference

M Milford, G Wyeth (2008). Single Camera Vision-Only SLAM on a Suburban Road Network, in proceedings of the IEEE international Conference on Robotics and Automation, Pasadena, United States.

PDF (1535K)

Fig 1. The original image (a) is converted to greyscale (b), then cropped and converted into a column intensity graph (c).

The race to learn: Spike timing and STDP can coordinate learning and recall in CA3

Christopher R. Nolan, Gordon Wyeth, Michael Milford, Janet Wiles.
The University of Queensland, Australia

ABSTRACT

The CA3 region of the hippocampus has long been proposed as an autoassociative network performing pattern completion on known inputs. The dentate gyrus (DG) region is often proposed as a network performing the complementary function of pattern separation. Neural models of pattern completion and separation generally designate explicit learning phases to encode new information and assume an ideal fixed threshold at which to stop learning new patterns and begin recalling known patterns. Memory systems are significantly more complex in practice, with the degree of memory recall depending on context-specific goals. Here, we present our spike-timing separation and completion (STSC) model of the entorhinal cortex (EC), DG, and CA3 network, ascribing to each region a role similar to that in existing models but adding a temporal dimension by using a spiking neural network. Simulation results demonstrate that (a) spike-timing dependent plasticity in the EC-CA3 synapses provides a pattern completion ability without recurrent CA3 connections, (b) the race between activation of CA3 cells via EC-CA3 synapses and activation of the same cells via DG-CA3 synapses distinguishes novel from known inputs, and (c) modulation of the EC-CA3 synapses adjusts the learned versus test input similarity required to evoke a direct CA3 response prior to any DG activity, thereby adjusting the pattern completion threshold. These mechanisms suggest that spike timing can arbitrate between learning and recall based on the novelty of each individual input, ensuring control of the learn-recall decision resides in the same subsystem as the learned memories themselves. The proposed modulatory signal does not override this decision but biases the system toward either learning or recall. The model provides an explanation for empirical observations that a reduction in novelty produces a corresponding reduction in the latency of responses in CA3 and CA1. © 2010 Wiley-Liss, Inc.

Reference

CR Nolan, G Wyeth, M Milford, J Wiles (2010) The race to learn: Spike timing and STDP can coordinate learning and recall in CA3, Hippocampus Vol(no):pages TBA.

Published Online: 15 Mar 2010

Full Text: PDF (Size: 980K) Supporting Information
Local copy (early view)

Fig 1. Pattern separation and completion.

Fig 2. How timing can signify the distinction between novel and familiar inputs.

Fig 3. Network architecture showing main regions, cell types, and the fan in-fan out

Fig 4. Modulation of the EC2-CA3 synapses affects the balance between pattern separation and completion.

A role for symmetric head-angular-velocity cells: Tuning the head-direction network.

Peter Stratton and Janet Wiles.
The University of Queensland

ABSTRACT

All adaptive systems require calibration. Computational models of the head direction (HD) system of the rat usually assume that the connections that maintain HD neuron activity are pre-wired and static. Ongoing activity in these models relies on precise attractor dynamics. It is currently unknown how such connections could be so precisely wired, and how accurate calibration is maintained in the face of ongoing noise and perturbation. A model of the HD system that uses symmetric head-angular-velocity (HAV) cells as a training signal shows that the HD system can learn to support stable firing patterns from poorly-performing, unstable starting conditions. The proposed calibration mechanism explains why symmetric HAV cells in the rat outnumber their asymmetric counterparts. The mechanism also conjectures that the efficacy of one synapse onto a postsynaptic cell can be controlled in part by activity received by that same cell on another synapse. If its existence in biological networks is confirmed, this mechanism will add significantly to our understanding of synaptic plasticity. 

Reference

Peter Stratton and Janet Wiles. A role for symmetric head-angular-velocity cells: Tuning the head-direction network. Frontiers in Systems Neuroscience, 2009 (COSYNE’09). (Abstract only)

PDF (270 K)

Automatic Calibration of a Spiking Head-Direction Network for Representing Robot Orientation

Peter Stratton*†, Michael Milford*†, Janet Wiles†, Gordon Wyeth† *Queensland Brain Institute and †School of Information Technology and Electrical Engineering, The University of Queensland

ABSTRACT

Calibration of movement tracking systems is a difficult problem faced by both animals and robots. The ability to continuously calibrate changing systems is essential for animals as they grow or are injured, and highly desirable for robot control or mapping systems due to the possibility of component wear, modification, damage and their deployment on varied robotic platforms. In this paper we use inspiration from the animal head direction tracking system to implement a self-calibrating, neurally-based robot orientation tracking system. Using real robot data we demonstrate how the system can remove tracking drift and learn to consistently track rotation over a large range of velocities. The neural tracking system provides the first steps towards a fully neural SLAM system with improved practical applicability through self-tuning and adaptation.

Reference

Peter Stratton, Michael Milford, Janet Wiles and Gordon Wyeth. Automatic Calibration of a Spiking Head-Direction Network for Representing Robot Orientation. In Proceedings of the Australasian Conference on Robotics and Automation, Sydney, Australia, 2009.

PDF (2,071 KB)

Comparing Kurtosis Score to Traditional Statistical Metrics for Characterizing the Structure in Neural Ensemble Activity.

Peter Stratton and Janet Wiles,
The University of Queensland

ABSTRACT

This study investigates the range of behaviors possible in ensembles of spiking neurons and the effect of their connectivity on ensemble dynamics utilizing a novel application of statistical measures and visualization techniques. One thousand spiking neurons were simulated, systematically varying the strength of excitation and inhibition, and the traditional measures of spike distributions – spike count, ISI-CV, and Fano factor – were compared. We also measured the kurtosis of the spike count distributions. Visualizations of these measures across the parameter spaces show a range of dynamic regimes, from simple uncorrelated spike trains (low connectivity) through intermediate levels of structure through to seizure-like activity. Like absolute spike counts, both ISI-CV and Fano factor were maximized for different types of seizure states. By contrast, kurtosis was maximized for intermediate regions, which from inspection of the spike raster plots exhibit nested oscillations and fine temporal dynamics. Brain regions exhibit nested oscillations during tasks that involve active attending, sensory processing and memory retrieval. We therefore propose that kurtosis is a useful addition to the statistical toolbox for identifying interesting structure in neuron ensemble activity.

Reference

P Stratton, J Wiles (2008). Comparing Kurtosis Score to Traditional Statistical Metrics for Characterizing the Structure in Neural Ensemble Activity. In M. Marinaro et al., editors, Dynamic Brain – from Neural Spikes to Behaviors, Springer LNCS V 5286, 2008, pp.115-122.

PDF (135K)

Different measures of spike train characteristics

Complex Spiking Models: A Role for Diffuse Thalamic Projections in Complex Cortical Activity

Peter Stratton, Janet Wiles   
The University of Queensland, School of Information Technology and Electrical Engineering, Queensland Brain Institute, Brisbane, Australia

Cortical activity exhibits complex, persistent self-sustained dynamics, which is hypothesised to support the brain’s sophisticated processing capabilities. Prior studies have shown how complex activity can be sustained for some time in spiking neural network models, but network activity in these models resembled high firing rate seizure which would eventually fail, leading to indefinite quiescence. We present a spiking network model of cortex innervated by diffuse thalamic projections, called the Complex Spiking Model (CSM). The model exhibits persistent, self-sustained, non-periodic, complex dynamics at low firing rates. Multiple network configurations were tested, systematically varying diffuse excitation from the thalamus, strength of the local cortical inhibition and excitation, neighbourhood diameters, synaptic efficacies and synaptic current time constants. Complex activity in all the network configurations depended strongly upon the strength of the diffuse excitation from the thalamus. We propose that diffuse thalamic projections to cortex facilitate complex cortical dynamics and are likely to be an important factor in the support of cognitive functions.

Self-sustained non-periodic activity in networks of spiking neurons: The contribution of local and long-range connections and dynamic synapses

Peter Stratton, Janet Wiles   
The University of Queensland, School of Information Technology and Electrical Engineering, Queensland Brain Institute, Brisbane, Australia

Cortical dynamics show self-sustained activity which is complex and non-periodic. Assemblies of neurons show transient coupling exhibiting both integration and segregation without entering a seizure state. Models to date have demonstrated these properties but have required external input to maintain activity. Here we propose a spiking network model that incorporates a novel combination of both local and long-range connectivity and dynamic synapses (which we call the LLDS network) and we present explorations of the network's micro and macro behaviour. At the micro level, the LLDS network exhibits self-sustained activity which is complex and non-periodic and shows transient coupling between assemblies in different network regions. At the macro level, the power spectrum of the derived EEG, calculated from the summed membrane potentials, shows a power-law-like distribution similar to that recorded from human EEG. We systematically explored parameter combinations to map the variety of behavioural regimes and found that network connectivity and synaptic mechanisms significantly impact the dynamics. The complex sustained behaviour occupies a transition region in parameter space between two types of non-complex activity state, a synchronised high firing rate regime, resembling seizure, for low connectivity, and repetitive activation of a single network assembly for high connectivity. Networks without synaptic dynamics show only transient complex behaviour. We conclude that local and long-range connectivity and short-term synaptic dynamics are together sufficient to support complex persistent activity. The ability to craft such persistent dynamics in a spiking network model creates new opportunities to study neural processing, learning, injury and disease in nervous systems.

Spike-time robotics: a rapid response circuit for a robot that seeks temporally varying stimuli.

Janet Wiles, David Ball, Scott Heath, Chris Nolan and Peter Stratton  The University of Queensland, Brisbane, Australia

In this paper we describe a spiking neural circuit inspired by the pyramidal-interneuron network gamma (PING) circuit modeled by Whittington and colleagues [Whittington, M.A., Traub, R.D., Kopell, N., Ermentrout, B., Buhl, E.H.: Inhibition-based rhythms: experimental and mathematical observations on network dynamics. International Journal of Psychophysiology 38 (2000) 315-336]. The spiking network controls a rat animat – a rodent-inspired robot that can autonomously explore and map its environment. We demonstrate how the neural controller directs the rat animat’s movement towards temporal stimuli of the appropriate frequency using an approach based on Braitenberg Vehicles. The circuit responds robustly (after four cycles) when first detecting a light pulsing at 1 Hz, and rapidly (after one-to-three cycles) when primed by recent experiences with the same frequency. This study is the first to demonstrate a biologically-inspired spike-based robot that is both robust and rapid in detecting and responding to temporal dynamics in the environment. It provides the basis for further studies of biologically-inspired spike-based robotics.

Inhibition Dominates the Early Phase of Up-States in the Basolateral Amygdala

Francois Windels,1* James W. Crane,2* and Pankaj Sah1 1Queensland Brain Institute, The University of Queensland, Brisbane, Queensland; and 2School of Biomedical Sciences, Charles Sturt University, Bathurst, New South Wales, Australia

Slow oscillations (
1 Hz) in neural activity occur during sleep and quiet wakefulness in both animals and humans. Single-cell recordings in cortical neurons have shown that these oscillations are driven by a combination of excitatory and inhibitory synaptic inputs. During up-states, although the ratio between them varies between cells, excitation and inhibition follow similar time courses. Neurons in the basolateral amygdala (BLA) also show slow oscillations between the resting membrane potential (down-state) and depolarized potentials (upstates). Delivery of footshock during the down-state fully reproduces up-states in these cells. Here we report that up-states in BLA principal neurons up-states begin with an excitatory drive that is rapidly (within 50 ms) overwhelmed by inhibitory input. This excess of inhibitory drive is short lasting (300–400 ms), after which up-states are maintained by a tight balance between excitation and inhibition. This initial large inhibitory input restricts action potential generation and reduces the firing frequency of these cells. These results indicate that, in contrast to cortical neurons, up-states in BLA neurons show an initial period of strong cortically driven feed-forward inhibition. For the remainder of the up-state, feedback inhibition then acts to balance excitatory input.

Spatial Cognition for Robots:
Robot Navigation from Biological Inspiration

Gordon Wyeth and Michael Milford. The University of Queensland.

The paper discusses robot navigation from biological inspiration. The authors sought to build a model of the rodent brain that is suitable for practical robot navigation. The core model, dubbed RatSLAM, has been demonstrated to have exactly the same advantages described earlier: it can build, maintain, and use maps simultaneously over extended periods of time and can construct maps of large and complex areas from very weak geometric information. The work contrasts with other efforts to embody models of rat brains in robots. The article describes the key elements of the known biology of the rat brain in relation to navigation and how the RatSLAM model captures the ideas from biology in a fashion suitable for implementation on a robotic platform. The paper then outline RatSLAM's performance in two difficult robot navigation challenges, demonstrating how a cognitive robotics approach to navigation can produce results that rival other state of the art approaches in robotics.