A method for determining a semantic concept associated with an audio signal captured using an audio sensor. A data processor is used to automatically analyze the audio signal using a plurality of semantic concept detectors to determine corresponding preliminary semantic concept detection values, each semantic concept detector being adapted to detect a particular semantic concept. The preliminary semantic concept detection values are analyzed using a joint likelihood model based on predetermined pair-wise likelihoods that particular pairs of semantic concepts co-occur to determine updated semantic concept detection values. One or more semantic concepts are determined based on the updated semantic concept detection values. The semantic concept detectors and the joint likelihood model are trained together with a joint training process using training audio signals, at least some of which are known to be associated with a plurality of semantic concepts.