Spread spectrum is the modulation technique which is recently used to embed the watermark in audio signals. However, the channel capacity of the spread spectrum watermarking has not been investigated in detail. In this paper the channel capacity analysis of the generalized spread spectrum watermarking (GSSW) scheme is presented, the theoretical capacity expression of the binary watermark is derived, and an approximate expression for the demand of engineering application is also presented. The optimum parameter values for channel capacity are obtained, and it is concluded that the maximum capacity of the GSSW scheme is larger than that of the traditional spread spectrum scheme because the host signal information is used in the embedded procedures.