[en] The capabilities of future HPGe arrays consisting of highly segmented detectors, like AGATA will depend heavily on the performance of γ-ray tracking. The most crucial component in the whole concept is the pulse shape analysis (PSA). The working principle of PSA is to compare the experimental signal shape with signals available from a basis set with known interaction locations. The efficiency of the tracking algorithm hinges on the ability of the PSA to reconstruct the interaction locations accurately, especially for multiple γ-interactions. Given the size of the arrays the PSA algorithm must be run in a real-time environment. A prerequisite to a successful PSA is an accurate knowledge of the detectors response. Making a full coincidence scan of a single AGATA detector, however takes between two and three months, which is too long to produce an experimental signal basis for all detector elements. A straight forward possibility is to use a precise simulation of the detector and to provide a basis of simulated signals. For this purpose the Java Agata Signal Simulation (JASS) was developed in the course of this thesis. The geometry of the detector is given with numerical precision and models describing the anisotropic mobilities of the charge carriers in germanium were taken from the literature. The pulse shapes of the transient and net-charge signals are calculated using weighting potentials on a finite grid. Special care was taken that the interpolation routine not only reproduces the weighting potentials precisely in the highly varying areas of the segment boundaries but also that its performance is independent of the location within the detector. Finally data from a coincidence scan and a pencil beam experiment were used to verify JASS. The experimental signals are reproduced accurately by the simulation. Pulse Shape Analysis (PSA) reconstructs the positions of the individual interactions and the corresponding energy deposits within the detector. This is accomplished by searching the simulated signal basis for the best agreement with the experimental signal. The particular challenge lies in the binomial growth of the search space making an intelligent search algorithm compulsory. In order to reduce the search space, the starting time t₀ for the pulse shapes can be determined independently by a neural network algorithm, developed in the scope of this work. The precision of 2 - 5ns(FWHM), which is far beyond the sampling time of the digitizers, directly influences the attainable position resolution. For the search of the positions the so-called 'Fully Informed Particle Swarm' (FIPS) was developed, implemented and has proofed to be very efficient. Depending on the number of interactions an accurate reconstruction of the positions is accomplished within several μs to a few ms. Data from a simulated (d, p) reaction in inverse kinematics, using a ⁴⁸Ti beam at an energy of 100 MeV, impinging on a deuterated titanium target were used to test the capabilities of the developed PSA algorithms in a realistic setting. In the ideal case of an extensive PSA an energy resolution of 2.8 keV (FWHM) for the 1382 keV line of ⁴⁹Ti results but this approach works only on the limited amount of data in which only a single segment has been hit. Selecting the same events the FIPS-PSA Algorithm achieves 3.3 keV with an average computation time of ∝ 0.9ms. The extensive grid search, by comparison takes 27ms. Including events with multiple hit segments increases the statistics roughly twofold and the resolution of FIPS-PSA does not deteriorate significantly at an average computing time of 2.2ms. (orig.)