Clinical gait analysis contributes massively to rehabilitation support and improvement of in-patient care. But there are still several unresolved shortcomings such as: the absence of medical records after hospital discharge, a lack of feedback to the subject, limited walking distances in intramural settings and affection of subjects through laboratory situation. A self-developed pair of instrumented shoe insoles, called eSHOE, fills this gap after hospital discharge and enables individual home-based monitoring and training. Motion and pressure sensors gather movement data directly on the (user's) feet, store them locally and transmit them wirelessly to a PC. In difference to current intramural and extramural gait analysis systems, the eSHOE hardware is fully embedded into a pair of orthopedic insoles. Furthermore, instead of delivering raw movement data, a combination of pattern recognition and feature extraction algorithms makes use of the multimodal sensor input and translates the motion data into standard gait parameters. Another important distinctive feature is that the measurement system, being completely integrated into a garment, enables the acquisition of movement data in an unobtrusive way. Thereby, the user or test subject is less influenced by the measurement process, because no cumbersome equipment has to be set up or attached to the body and the subject is not in an in the spotlight sort of situation. Additionally, there are no spatial restrictions, so subjects can be analyzed on walking tracks of any length and with different underground structures. As a first step in the investigation whether eSHOE provides useful results, the reliability of the pattern detection and feature extraction results have been compared to actual number of gait cycles , which were manually counted in the raw data. The same was done for the number of basic features, such as initial contact (IC) and last contact (LC). The accuracy of the subsequently calculated temporal gait parameters was evaluated against the reference system GAITRite® in the course of a clinical pilot study. Those parameters are stride time (STR), stance time (STA), swing time (SWI), step time (STE) and double support time (DST). Eleven hip fracture patients (78.4 ± 7.7 years) and twelve healthy subjects (40.8 ± 9.1 years) were included in these trials. All subjects performed three measurements at a comfortable walking speed over eight meters, including the six-meter-long GAITRite® mat. These validation measurements were also used to investigate, whether it is possible to distinguish between eSHOE data collected from healthy subjects (CTRL) and patients (PAT). Furthermore, a comparison between healthy and injury-affected leg, within PAT, has been carried out. For these purposes xi two sets of non-parametric statistical tests have been chosen. The differences between the groups have been evaluated using the median test, the Mann-Whitney U-test and the Kolmogoroff-Smirnov test. Measurements of healthy and affected leg have been compared by the sign test and the sign rank test. In the course of the same pilot study it was attempted to document the course of each patient's therapy success by means of the development of his or her gait parameters that can be calculated from the eSHOE data. For that purpose each patient was asked to perform three repetitions of the 10-meter walk test (10MWT), a standardized mobility assessment, while wearing a pair of eSHOE insoles, once a week during their (three- to five-week) stay. Non-parametric measures were used to determine differences between healthy and affected leg. Regarding therapy progress, data were evaluated by using descriptive statistics. Stride length (STL) and total distance estimation have also been evaluated in another (non-clinical) study. Several already existing approaches for distance estimation via inertial sensor signals, based on inertial navigation and dead reckoning, have been consulted. A specially designed method, which utilizes certain aspects from those methods, has been developed. It involves modeling of the random stride-to-stride bias offset and the random linear bias drift in combination with a well-timed zero-velocity-update Kalman filter. Three different parameter sets for this method have been implemented and validation against manual means (measurement tape) and the optical motion capturing system VICON® has been performed. The extraction of sensor- and axis-specific patterns resulted in detection accuracies from 91.4 to 99.8 %. Major gait events, like initial and last contact, could be extracted with an accuracy of 97.6 % (mean left) and 98.6 % (mean right). Six temporal gait parameters were extracted from a total of 347 gait cycles. Agreement with the reference system GAITRite® was analyzed via scatter-plots, histograms and Bland-Altman plots. In the patient group the average differences between eSHOE and GAITRite® range from -0.046 to 0.045 s and in the healthy group from -0.029 to 0.029 s. The best parameterset for stride length estimation managed to achieve an accuracy of -3.0 ± 2.2 %, applied to an average stride length of 1.3 m this results in -3.9 ± 2.9 cm/stride. For the total distance estimation at 50 m this parameter-set exhibited an (average) error of -4.5±0.7 % (-2.26 ± 0.35 m) and -2.0 ± 2.4 % (-0.17 ± 0.21 m) at 8.7 m. Significant differences between CTRL and PAT could be identified visually and confirmed via statistical tests in both eSHOE and GAITRite® data. The comparison of healthy and affected leg among patients, regarding eSHOE data, presented equal results for stride time, step time and terminal double support time and showed statistical significant differences in stance time, swing time and double support time. The evaluation of the patient group's gait parameters' progression during the stay in geriatric care revealed the same statistical significant differences between healthy and affected. Therefore, the evaluation of therapy progress was performed for each leg separately in those parameters showing differences. The inspection of the course of the external reference data, the 10MWT duration, already reveals that there is no consistently identifiable progress in the entirety of all test subjects. There is a clear reduction of duration's median and mean after the first measurement day, but afterwards the results stagnate. On the sixth and last day there is even a slight increase. The gait parameters' median and mean show similar courses. Even the changes after the first day are only detectable in some parameters. However, the distributions of all parameters, standard deviation (SD), interquartile range (IQR) and total range (TR), do change. From day one to day six, standard deviation (SD) is reduced by 82%, interquartile range (IQR) by 72% and total range (TR) by 84%. The differences between the means of healthy and affected leg in both, stance and swing phase duration, are getting smaller over time. While a ground contact (stance time) on the healthy foot lasts 0.091 s longer than on the injured leg on the first day, it is only 0.015 s longer on the last day. That equals a reduction in difference of 84%. The mean swing time of the healthy foot on the first day is 0.119 s shorter and 0.017 s on the last day, reducing the difference by 86%. In case of the double support time there was also a global reduction in the gap between both legs by 75% from -0.071 s to -0.018 s. In reference to the validation results, it can be concluded that eSHOE delivers adequately accurate results regarding all temporal gait parameters. The discrepancy between the CTRL and the PAT group indicate, that the eSHOE algorithms work less effective and accurate with motion data from patients. This may root in the fact that the raw data from patients is more erratic and, therefore, more difficult to process for the algorithms. Nevertheless, deviations in the are of ±46 ms are still acceptable. For the stride length and total distance estimation it seems that the given combination of hardware and calculation methods work more accurate over long distances than they do on shorter tracks, such as a human stride. Higher sample rate, new inertial measurement unit models, with better signal quality, and (mathematically) more sophisticated error compensation methods my provide better results in the future. Therapy progress could not be determined directly by the detection of clear change of any parameters median or mean during the stationary stay. However, a narrowing of the distributions, via consistent reductions of SD, IQR and TR, was present in all extracted gait parameters. This indicates, that eSHOE was at least able to detect a decrease in inter-personal gait variability. It is possible that the information about the actual progress is lost among the very broad distributions, especially occurring in the first half of the hospital stay. The individual analyses of certain test subjects, where e.g. a reduction in stride time was detectable, supports this hypothesis. Possible causes are assumed to be the small size and the heterogeneous composition of the sample. It is very likely, that the different usage of walking aids in each individual case at different times influenced the development of the gait parameters and caused the broad distributions. eSHOE's accuracy doesn't quite match the stationary systems', but it allows the distinction between healthy subjects and persons with unilateral injuries of their lower extremities. Measurements can be performed with less effort and the system is more cost-efficient. It is as accurate as other mobile gait analysis systems, with the difference that it is completely integrated in a pair of shoe insoles. Thus, eSHOE is well-suited for unobtrusive, long-term measurement of motion data with only minimal influence on it's wearer. Soon, the steady growing number of older people in the population and the related increasing incidences of chronic diseases will call for solutions to ensure long-term success of therapy and rehabilitation success. This need could be met by the envisaged further development of eSHOE into a rehab@home system, that supports the success of therapy measures. The currently booming 'quantified self' trend provides additional opportunities for lifestyle application scenarios, which are not limited to a specific age group.