A method of processing an audio signal, and which includes a receiving object information including object level information indicating a level of objects, object gain information and object correlation information; receiving preset presence information and preset number information; receiving preset information based on the preset presence information and the preset number information by an audio decoding apparatus, the preset information including preset metadata representing an attribute mode of the preset information, preset rendering data and preset type information indicating whether the preset rendering data is in a matrix type or not obtaining a preset matrix from the preset rendering data by the audio decoding apparatus, and generating a processed audio signal by controlling a gain or panning of the at least one object included in the audio signal based on the downmix processing information.