Albert Kern

Analog electronic cochlea with mammalian hearing characteristics

Systems close to bifurcations can be used as small-signal amplifiers. Biophysical measurements suggest that the active amplifiers present in the mammalian cochlea are systems close to a Hopf bifurcation. The pure tone and transient signal output of our electronic hearing sensor based on this observation provides output that is fully compatible with the electrophysiological data from the mammalian cochlea. In particular, it reproduces all salient nonlinear effects displayed by the cochlea.

Sequential clustering: tracking down the most natural clusters

Sequential superparamagnetic clustering (SSC) is a substantial extension of the superparamagnetic clustering approach (SC). We demonstrate that the novel method is able to master the important problem of inhomogeneous classes in the feature space. By fully exploiting the non-parametric properties of SC, the method is able to find the natural clusters even if they are highly different in shape and density. In such situations, concurrent methods normally fail. We present the results from a fully automated implementation of SSC (applications to chemical data and visual scene analysis) and provide analytical evidence of why the method works.

Sequential superparamagnetic clustering for unbiased classification of high-dimensional chemical data.

For the clustering of chemical structures that are described by the Similog, ISIS count, and ISIS binary fingerprints, we propose a sequential superparamagnetic clustering approach. To appropriately handle nonbinary feature keys, we introduce an extension of the binary Tanimoto similarity measure. In our applications, data sets composed of structures from seven chemically distinct compound classes are evaluated and correctly clustered. The comparison, with results from leading methods, indicates the superiority of our sequential superparamagnetic clustering approach.

A generalization of the van-der-Pol oscillator underlies active signal amplification in Drosophila hearing

The antennal hearing organs of the fruit fly Drosophila melanogaster boost their sensitivity by an active mechanical process that, analogous to the cochlear amplifier of vertebrates, resides in the motility of mechanosensory cells. This process nonlinearly improves the sensitivity of hearing and occasionally gives rise to self-sustained oscillations in the absence of sound. Time series analysis of self-sustained oscillations now unveils that the underlying dynamical system is well described by a generalization of the van-der-Pol oscillator. From the dynamic equations, the underlying amplification dynamics can explicitly be derived. According to the model, oscillations emerge from a combination of negative damping, which reflects active amplification, and a nonlinear restoring force that dictates the amplitude of the oscillations. Hence, active amplification in fly hearing seems to rely on the negative damping mechanism initially proposed for the cochlear amplifier of vertebrates.

Collective bursting in layer IV. Synchronization by small thalamic inputs and recurrent connections.

Layer IV is believed to be the cortical signal amplifier, for example, of thalamic signals. A previous spiny stellate recurrent network model of this layer is made more realistic by the addition of inhibitory basket neurons. We study the persistence and characteristics of previously observed collective firing behavior, and investigate what additional features would need to be implemented to generate in vivo type neuronal firing. It is shown that neuronal activity is only coarsely synchronized within the network. By applying methods of noise-cleaning, it emerges that the firing of individual neurons is of low-dimensional hyperchaotic nature, as found in the analysis of measured cat in vivo spike trains. In order to reproduce in vivo firing patterns, it is sufficient to have time-varying thalamic input. Conclusions from low-dimensional hyperchaotic behavior of network-embedded neurons are drawn. We interpret observed in vivo pattern-sharpening features of stimuli and outline possible connections to epilepsy. From our results, it follows that emergent global behavior is likely to be the result of the interaction between comparably simple neuronal components, driven by input specificity.

Biophysical Parameters Modification Could Overcome Essential Hearing Gaps

A majority of hearing defects are due to malfunction of the outer hair cells (OHCs), those cells within the mammalian hearing sensor (the cochlea) that provide an active amplification of the incoming signal. Malformation of the hearing sensor, ototoxic drugs, acoustical trauma, infections, or the effect of aging affect often a whole frequency interval, which leads to a substantial loss of speech intelligibility. Using an energy-based biophysical model of the passive cochlea, we obtain an explicit description of the dependence of the tonotopic map on the biophysical parameters of the cochlea. Our findings indicate the possibility that by suitable local modifications of the biophysical parameters by microsurgery, even very salient gaps of the tonotopic map could be bridged.

Shiner?Davison?Landsberg complexity revisited

Shiner, Davison and Landsberg have recently proposed a measure of complexity that has become the subject of an intense debate. We show that using the framework of the thermodynamic formalism, the properties and shortcomings of this measure—over-universality and a trivial implementation of the temperature dependence—can be interpreted and elucidated in a coherent way. Moreover, we show how the SDL approach can be refined to nullify these critiques. Results of the logistic parabola family demonstrate the improved behaviour of the modified SDL measure of complexity. For the tent map family, an interesting linear dependence of the modified measure as a function of the asymmetry is observed.

Analysis of the "Sonar Hopf" Cochlea

The “Sonar Hopf” cochlea is a recently much advertised engineering design of an auditory sensor. We analyze this approach based on a recent description by its inventors Hamilton, Tapson, Rapson, Jin, and van Schaik, in which they exhibit the “Sonar Hopf” model, its analysis and the corresponding hardware in detail. We identify problems in the theoretical formulation of the model and critically examine the claimed coherence between the described model, the measurements from the implemented hardware, and biological data.

Principles and typical computational limitations of sparse speaker separation based on deterministic speech features

The separation of mixed auditory signals into their sources is an eminent neuroscience and engineering challenge. We reveal the principles underlying a deterministic, neural network–like solution to this problem. This approach is orthogonal to ICA/PCA that views the signal constituents as independent realizations of random processes. We demonstrate exemplarily that in the absence of salient frequency modulations, the decomposition of speech signals into local cosine packets allows for a sparse, noise-robust speaker separation. As the main result, we present analytical limitations inherent in the approach, where we propose strategies of how to deal with this situation. Our results offer new perspectives toward efficient noise cleaning and auditory signal separation and provide a new perspective of how the brain might achieve these tasks.

Pattern detection in noisy signals

Methods for detecting patterns in noisy signals are often template based. As a consequence, a priori selections of potential pattern structures have to be made. To avoid this shortcoming, we propose a novel statistical approach based on the correlation integral. The method significantly reduces the set of appropriate templates, and also works under noisy conditions.