It is important to maintain IAQ as well as to save energy in buildings. Demand controlled ventilation has been proposed to control ventilation rate according to the occupancy in a building. The CO₂ concentration is widely used as an indicator of the indoor pollution and often changes with time on account of the number of occupants. The objective of this paper is to investigate the performance for estimating occupancy when we use humidity data in addition to carbon dioxide concentration data using dynamic neural network method. Experiments have been conducted to include seasonal effect and results have been analyzed with and without humidity data. Absolute humidity has been found not to improve the estimation performance because of the condensation and evaporation of moisture especially in summer when air-conditioner is operating.