Steganography is considered as one of the widely used methods for hiding information; it hides secret data in digital cover data without clear suspicion. This work focus on using two transform based methods for hiding secret data. First, the data will converted to binary form, then two predefine threshold values are used to categorize the audio blocks. The first threshold is used to decide whether the block is voiced/unvoiced, and the second threshold is used to decide which transform will be used on the voiced blocks only (i.e., DCT or Wavelet). Second, the audio signal will be transformed by the chosen one of transforms; DCT transform will applied on voiced blocks that have energy less than second threshold value, and wavelet transform will be applied on voiced blocks that have energy grater than or equal to the second threshold value. In the third stage the embedding on secret data on voiced blocks is applied. The performance of the proposed hybrid system was tested by using some fidelity measures (MSE, MAE, and PSNR) to measure the rate of error, the ratio of corrected retrieved bits and the hiding rate, the later is computed to assess the embedding power of system. Also, the effects of some control parameters on the system performance were investigated to assist user to correctly choose the proper values of the system parameters, some of these parameter is the threshold value which classify the blocks to voice/unvoiced, and our goal is to hide in voiced blocks, this will provide high hiding rate for the system.