Chaotic Masking Of Speech Signals And Its Blind Extraction Techniques

With the layout of social development's global strategy, the development tendency of information technology increases rapidly, having penetrated into all areas, such as military security and daily life. Information transmission problem has been put forward by some scholars. Degree of communication confidentiality can help ensure the secure of information transmission. With features such as sensitivity to the initial value, randomness and similar noise wide frequency spectrum, chaotic signal is difficult to be deciphered., accessing to a wide range of applications in many fields, providing an effective carrier for signal's secure transmission. Speech signal is the main medium of human information transmission, the security transfer is of great significance. In this thesis, chaotic signals works as a carrier. This thesis studies on chaotic hiding secret transmission method for speech signal to achieve effective blind source separation or blind extraction of speech signals having important theoretical and practical value.Hide speech signal within the chaotic carrier signal and realize covering based on chaotic obscure technology theory. Select multidimensional chaotic systems as the carrier which have complexed dynamic characteristics and using time-frequency analysis for speech signal and chaotic signal. Because speech signals are small signals, and chaotic signal's energy level is relatively high, so it can effectively achieve chaotic hiding, and goal of secure communication. Simulations in this paper starts from chaotic hiding speech blind signal separation in ideal positive definite systems and studies on realizability and universality in a chaotic system using fast algorithm for independent component analysis(Fast ICA). Take actual communication environment into account and consider the additive noise and the number of receive antennas smaller than the number of transmit antennas, namely blind extraction of speech signals in under-determined system. Because chaotic signals are of non-stationarity and speech signal does not meet the sparsity, use the empirical mode decomposition algorithm and establish virtual receiver array to turn under-determined system into a definite system addressing morbid problem. Use Fast ICA algorithm realizes blind extraction of interested speech in under-determined system. Simulations have good results in different dimensions systems belonging to different dimensions verifying the validity and universality.