Wireless patient monitoring using wearable sensors is a promising application. This paper provides stochastic channel models for wireless body area network (WBAN) on the human body. Parameters of the channel models are extracted from measured channel transfer functions (CTFs) in a hospital room. Measured frequency bands are selected so as to include permissible bands for WBAN; ultra wideband (UWB), the industry, science and medical (ISM) bands, and wireless medical telemetry system (WMTS) bands. As channel models, both a path loss model and a power delay profile (PDP) model are considered. But, even though path loss models are derived for the all frequency bands, PDP model is only for the UWB band due to the highly frequency selectiveness of UWB channels. The parameters extracted from the measurement results are summarized for each channel model.