Audio sound data (compressed data) read by an optical pickup is temporarily held in a memory and the audio sound data is expanded to be returned to its original size. At the time of usual reproduction, since the difference in data transfer rates of reading by optical pickup and data expansion is absorbed, writing and reading of audio sound data are alternately performed using part of the memory capacity, while audio sound data is held by a fixed amount for data protection upon impact occurrence utilizing the remaining memory capacity. At the time of track jump, for example, if the moving distance of a head is long, all the memory capacity is used for data protection. If the moving distance of the head is short, part of the memory capacity is allocated according to the moving time, and the remaining memory capacity is used for data protection upon impact occurrence.