Characteristic Delay and Characteristic Phase in Broadband Interaural Time Difference Tuning Curves Arising by Across-Frequency Integration
- Entstehung von charakteristischem Delay und charakteristischen Phase in Breitband ITD Tuning-Kurven mittels Frequenzintegration
Lehmann, Jessica Frida-Anita; Führ, Hartmut (Thesis advisor); Wagner, Hermann (Thesis advisor)
Aachen : Publikationsserver der RWTH Aachen University (2015)
Dissertation / PhD Thesis
Aachen, Techn. Hochsch., Diss., 2015
An important quantity in the sound localization processing, used by many animals, is the interaural time difference (ITD), the time shift between the acoustic signals reaching the two ears. The first step in the representation of ITD, the detection step, may be described by the place theory of Jeffress. Binaural neurons are modeled as coincidence detectors which respond preferable to their best ITD. After the detection stage, the neural response is remodeled which involves across-frequency integration. Tonal ITD tuning curves of 'integrator neurons' collected for different frequencies show a common relative discharge rate at a single delay which occur either at the best ITD for all frequencies or at a common point on the slope. Consequently, the ITD tuning curves may be described in terms of frequency-dependent component, the characteristic phase (CP), and frequency-independent component, the characteristic delay (CD). Assuming that a broadly tuned neuron linearly sums up convergent input from narrowly tuned coincidence detector neurons, each characterized by a different best frequency and a best ITD, we developed four different algorithms for the estimation of the CD and the CP from either the frequency tuning curves or from broadband ITD tuning curves. The systematic dependence of the best ITD on the best frequency is responsible for the emergence of the CP and the CD. We used numerical experiments to study properties of the algorithms and to compare the algorithms. Each algorithm is applied to artificially generated tuning curves with predetermined CD and CP values which take into account the influence of noise as well as the frequency band dependent response behavior of a neuron, as it is observed in experimental data. The estimation performance of all algorithms was bandwidth dependent, and to a smaller degree dependent on the position of the frequency band. Compared to all other developed estimation algorithms, the MSE fit algorithm resulted in smaller mean estimation errors. In terms of overall performance, it was superior to all other estimation algorithms, especially for the estimation of the CP. The algorithms are based on certain model assumptions regarding the emergence of the CD and the CP as a result of frequency integration. We present a simple two-layer neuronal model that corroborates this assumption and provides further evidence that the estimation algorithms are sound. The MSE fit algorithm is systematically applied to electrophysiological data of the barn owl and also to sample data for mammalian. The resulting CD and CP values were in great accordance with previous findings, indicating that the MSE fit algorithm is also valid for the estimation of the CD and the CP for other species.