Structured Spectral Compression for Low-Bitrate Secure Speech Communications

With the rapid advancement of information technology and network science, speech communication has assumed a pivotal role in our daily lives, commercial activities, and even military operations. However, the widespread application of speech data has also exacerbated the concern on the privacy disclosure. To ensure the personal privacy carried in the speech data while simultaneously conserving bandwidth resources and enhancing the efficiency and quality of speech communications, this talk focuses on the low-bitrate secure speech communications based on the structured spectral compression. At the transmitter side, the Mel spectrogram matrix of the speech signal is encoded into a sparse matrix through the waveform segmentation and high-order matrix sparsification based on the compressive sensing. To further reduce the number of transmitted bits, the quantization is performed on the compressed spectral data. At the receiver side, the speech is restored based on the Mel-spectrum reconstruction through separating the Mel-spectrum dimensions and sparse vectors. To improve the restored speech quality, the noise is reduced based on the singular spectrum analysis. Our experiments reveal that the structured coding method proposed herein boasts a time complexity of merely O(n) and requires a mere 10 bits of memory, both significantly smaller than those of traditional encryption algorithms. Furthermore, compared to conventional compression techniques, our method reduces the compression time by over 90%, substantially boosting the efficiency. It necessitates a coding rate of only 7.1kbps, approaching the 6.3kbps required by G.711 without the privacy disclosure.