Wearable accelerometers provide an objective measure of human physical activity. We extract meaningful features from the raw accelerometry data and develop and evaluate a classification method for the detection of walking and its subclasses, i.e. walking, descending, and ascending stairs. Our methodology is tested on a sample of 32 adults for whom we extracted features based on the Fourier and wavelet transforms. We build subject and group level classification models utilizing a tree-based methodology. We evaluate the effects of sensor location and tuning parameters on the classification accuracy. In the group level classification setting, we propose a robust feature inter-subject normalization and evaluate its performance compared to unnormalized data. The classification accuracy on the subject level was on average 87.6%, with the ankle-worn accelerometers showing the best performance with an average accuracy 90.5%. Group level classification accuracy using the normalized features was 80.2% compared to 72.3% for the unnormalized features. A framework is provided for better use and feature extraction from the raw accelerometry data to differentiate between walking modalities.