A cell expressing an olfactory receptor (OR) exhibits excellent odorant detection ability and thus is widely applied in odor biosensors. Most of those biosensors, however, could detect only liquid-phase nonchanging single-component odorants. In this paper, we raised up an odor biosensor for the active tracking of temporally changing gas-phase odor mixture by an array of cells expressing ORs. A thin stable liquid film covered the cell, thus allowing gas-phase odorants to penetrate. The online image processing generated individual cell brightness data which were used to compute the biosensor response. Based on the obtained responses, we adjusted the known odor components to be similar with the unknown odor. The function of our biosensor was validated by tracking the variable single-component odorant or the binary odor mixture. The influence from the sensor drift could be overcome by comparing the adjacent unknown and known odor responses. In the odor mixture quantification, adding the OR label to mixed cells and then quantifying separately (named as the pre-label method) was more efficient, while directly using the cell response pattern (named as the label-free method) was still capable even if the OR odor had cross-sensitivity.