In this paper, we present an approach for data hiding within speech signals based on Hash and Hilbert Transform (HT). The secret data is firstly pre-processed by Hash to enhance its security. Then, exploiting the orthogonality of HT and insensitivity of human perception to the phase of speech, we embed processed secret data into a speech signal by using the HT. We propose in this paper both blind and non-blind methods to extract embedded data. Experiment results have shown that the proposed scheme demonstrates transparent, secure and robust data hiding performances.