Research On Technology For Encrypted Speech Via GSM

As there are lots of security flaws in the public mobile communicationsnetwork, there is a need of end-to-end voice secure communication which inducessome problems. For that, this paper researches on the voice service channel basedscheme of end-to-end voice secure communication. To transmit encrypted dataagainst vocoder’s compression in the scheme, this paper analyses the existing methods,sums up the functions and characteristics of speech-like modulation anddemodulation, and then lays its focus on the technology of speech-like modulationand demodulation from following two aspects.First of all, by modeling analysis of vocoder, this paper mensurates thefrequency response of "black box" channel system composed by GSM vocoder,researches and implements a speech-like modulation and demodulation algorithmbased on channel model. This algorithm uses OFDM technology to carry data on themodulated signal in which the subcarrier’s frequency band is limited to theapproximate band-pass range of the ’black box" system, so that the signal can passthrough the vocoder channel with less distortion and be demodulated exactly,meanwhile channel coding and sequence encryption has been used to reduce the errorrate and prevent the error diffusion. Then data is transmitted against vocoder’scompression. Simulation results show that this algorithm has lower bit error rate andhigher modulation rate, so it can better ensure real time communication.Second, through researching vocoder’s features on voice coding and decoding,the vocoder distortion can be divide into inner-distortion and inter-distortion, aspeech-like modulation and demodulation algorithm based on two codebooks isproposed to resist the two types of distortion, and the encrypted voice was modulatedby different codebooks. To reduce the inner-distortion, codebook symbols are chosenfrom TIMIT which is a data base of human speech, and a set of speech symbols weregenerated with a high adaptability to vocoder by replacing the most error causingsymbol, meanwhile to reduce the inter-distortion, the codebooks were generated bygradually combined symbols with which has the maximum Hamming distance, sothat to maximize channel capacity. Simulation results show that the voice modulatedby double codebooks has a dynamic spectral envelope which and avoid the VoiceActivity Detection, and this algorithm has higher modulation rate than the methodbased on formant model, when reaching the condition of bit error rate that cantransmit speech.