This paper presents a framework based on Gaussian Processes for assessing cross channel consensus in Body Sensor Network (BSN) data. Cross channel consensus can be observed by measuring the prediction error of one channel given the others, which could help in predicting missing data, correcting for noisy channels, or learning relationships between sensor channels over time. The method is evaluated with activities of daily living experiments with sensing data including heart rate, respiration and activity levels. The acquired prediction rates indicate the potential practical value of the technique for home-monitoring of chronically ill patients.