A proposed analog speech scrambler based on parallel structure of wavelet transforms

Abstract

The main aim of this paper is to modeling and simulating a proposed analog speech scrambling system for analog speech signals, based on parallel structure of wavelet transformations. The proposed speech scrambling system depends on the parallel structure of using two different types of wavelets with the same decomposition levels. In this system, Db1 wavelet along with Haar wavelet, Db2 wavelet along with Sym2 wavelet and Db4 wavelet along with Sym4 wavelet for the same level are implemented. The performance evaluation process of the proposed system was taken into consideration via the using of noise power levels (5, 15 and 25 dB) in the simulated communication channel. The system performance was tested using the subjective test, by plying back the original, scrambled, and descrambled signal to a number of listeners in order to estimate the intelligibility of the played signals, and objective test, using the (linear prediction coding, cepstral distance and segmental signal to noise ratio). The results of this work is promising, and they can be considered as a backbone for a future programs, since for the first time the concept of parallel structure for transformation was introduced in analog speech scrambling techniques based on the transform domain philosophy. This approach provides more security for the resulted scrambled speech signal in comparison with the conventional scrambling approach using only a single orthogonal transformation like FFT, DCT, or Walsh Hadmard transform (WHT), and the DWT in classical structure.