Nonlinear Dynamics Principles of Hearing ; Pitch Sensation; Source Separation; Hearing Sensor Periphery; Cochlea Implants; Hearing Robots
Stoop Ruedi, Kanders Karlis, Lorimer Tom, Held Jenny, Albert Carlo (2016), Big data naturally rescaled, in Chaos, Solitons and Fractals
Gomez Florian, Lorimer Tom, Stoop Ruedi (2016), Signal-coupled subthreshold Hopf-type systems show a sharpened collective response, in Physical Review Letters
, (116), 108101.
Lorimer Tom, Gomez Florian, Stoop Ruedi (2015), Two universal physical principles shape the topological statistics of real-world networks, in Scientific Reports
, (5), 12353.
Ferrari Fabiano A.S., Viana Ricardo, Gomez Florian, Lorimer Tom, Stoop Ruedi (2015), Macroscopic bursting in physiological networks: node or network property?, in New Journal of Physics
, (17), 055024.
Lorimer Tom, Gomez Florian, Stoop Ruedi (2015), Mammalian cochlea as a physics guided evolution-optimized hearing sensor, in Scientific Reports
, (5), 12353.
Gomez Florian, Stoop Ruedi (2014), Mammalian pitch sensation shaped by the cochlear fluid, in Nature Physics
, (10), 510.
Andres D.S., Gomez F., Cerquetti D., Merello M. (2014), A Hierarchical Coding-Window Model of Parkinson's Disease, in Nonlinear Dynamics of Electronic Systems 2014
, Albena, BulgariaSpringer, Cham, Switzerland.
Gomez F., Lorimer T., Stoop R. (2014), Complex Networks of Harmonic Structure in Classical Music, in Nonlinear Dynamics of Electronic Systems 2014
, Alben, BulgariaSpringer, Cham, Switzerland.
Lorimer T., Gomez F., Stoop R. (2014), Deviation from Criticality in Functional Biological Networks, in Nonlinear Dynamics of Electronic Systems 2014
, Albena, BulgariaSpringer CCIS 438, Cham, Switzerland.
Gomez F., Saase V., Buchheim N., Stoop R. (2014), How the Ear Tunes In to Sounds: A Physics Approach, in Physical Review Applied
, (1), 014003.
Andres D.S., Gomez F., Ferrari F.A.S, Cerquetti D., Merello M., Viana R., Stoop R. (2014), Multiple-time-scale framework for understanding the progression of Parkinson's disease, in Physical Review E
, (90), 062709.
Gomez F., Stoop R.L., Stoop R. (2014), Universal dynamical properties preclude standard clustering in a large class of biochemical data, in Bioinformatics
, (30), 2486.
Martignoli Stefan, Gomez Florian, Stoop Ruedi (2013), Pitch sensation involves stochastic resonance, in Scientific Reports
, (4), 2676.
Stoop Ruedi, Gomez Florian, Auditory power-law activation avalanches exhibit a fundamental computational ground state, in Physical Review Letters
, accepted 20. Mai 2016.
The design of intelligent machines that not only communicate semantically with people, but also correctly interpret semantic contents in the emotional context, is pivotal for the development of ICT technologies and unavoidable for a beneficial use in society. This necessity is, in particular, a main bottleneck in the application of artificial intelligence to hearing. To resolve this problem, a number of diverse scientific and technical issues must be mastered. Profound knowledge must be acquired of how sound information is generated, monitored, and interpreted by the nervous system. Uncovering the general physical and the biophysical principles of sound information generation and decoding, fortunately, embodies the ability to implement these insights within lean and real-time capable technical frameworks. Presently, this level of knowledge is not achieved: A physics definition of the most relevant attributes of sound, pitch and timbre, is still essentially missing. The current common approach towards this goal is either by classical linear signal processing or by very detailed implementations of the biophysical templates of often unknown function. Due to the inherent complexity of both approaches, they have not led to efficient implementations of the functionalities required by artificial intelligence. As an example, the mammalian cochlea is fundamentally divergent in construction, functionality and performance from the essentially passive microphone systems used currently, designed only to convert sound pressure linearly into an electric voltage. The biological ear is a strongly nonlinear dynamical apparatus that not only detects sounds, but also provides great signal preprocessing. Nonetheless, its inherent nonlinear dynamics construction principles are relatively simple.We therefore propose a radically new paradigm. This paradigm is based on a simpler modelling level using the nonlinear dynamics approach, to provide an optimal solution to both the problem of understanding the auditory system at a fundamental level, and to obtain therefrom technological implementations with superior processing capabilities. Auditory processing in biological systems makes heavy use of collective, self-organizing processes, which are central manifestations of nonlinear dynamics. This key element is exploited by nature, but is missing in the complex artificial implementations of auditory systems. Our specific hypothesis is that most of the important features of hearing can be understood and be coded in terms of robust properties of moderately low dimensional dynamical systems. By using this approach for the cochlea, it has become possible to explain a large number of human sound perception phenomena from purely physical grounds. As probably the most important one, the perception of pitch has been explained, including, e.g. the frequency selectivity, nonlinear amplification, combination tone generation, suppression of neighbouring tones, the missing fundamental and the various pitch shift phenomena. Of interest is that all these phenomena are linked, where the link can easily be captured and exhibited in terms of notions of nonlinear dynamical systems. In our project we will apply this approach to understand unresolved phenomena of human and animal sound perception. We will interpret and study the biological peripheral circuitry (afferent and efferent connections to the cochlea) from a dynamical system point of view. The gained dynamical systems templates provide us with simple technical implementations that we will use to gain further insight into the biology blueprint’s nature, by comparing data from the implemented models with biological data. The paradigm will be cast into an expandable integrated hardware/software platform, that will reproduce the most salient aspects of natural hearing. Being fully accessible on all levels, the framework will be used at different stages of its completion not only as a test-bench for accepted concepts, but also as a predictive tool for unresolved aspects, and to explore the dynamical basis of less understood auditory features, paving the way for their implementation in working applications. We will primarily focus on the pitch as the acoustic feature, to be reproduced by the artificial cochlea and to be extracted by the constructed auditory platform. Pitch is not a fundamental physical quantity, but rather a program of extraction from a wave signal of arbitrary complexity; its evaluation demands both the ??correct generation of the pitch substrate in the cochlea, as well as the correct extraction of the feature. Pitch, moreover, is the fundamental quality used for efficient mono-channel voice separation and noise suppression.