We investigate the design space of flexible, textile capacitive sensors for applications in human activity recognition. In a previous paper, we showed that conductive textile patches can be used to measure capacitance of the human body and could reveal information about a broad range of activities. In this paper, we systematically investigate how different design parameters such as electrode size, electric field frequency, and the concrete analog circuit design influence sensor performance. To this end, we combine FEM electric field simulations, circuit analysis, and measurements. We illustrate the performance of sensor systems that implemented according to the design guidelines that we derived. Results from four typical activity recognition scenarios were considered, including heart rate and breathing rate monitoring, hand gesture recognition, swallowing monitoring, and gait analysis.