A Computationally Efficient CELP Codec with Stochastic Vectorquantisation of LPC Parameters
R. Salami, Lajos Hanzo, D.G. Appleby
Abstract
Open-access reader
R. Salami, Lajos Hanzo, D.G. Appleby
Abstract
Open-access reader
In this contribution, computationally efficient CELP structures are proposed for low-complexity speech coding below 6.4Kbits/s. We also report on a new approach for low bit rate speech coding using binary excitation pulses, which delivers the quality of the CELP codec without the excitation codebook storage, and with a very simple excitation determination procedure. Furthermore, we propose a switched-adaptive stochastic vector quantisation scheme for quantising the line spectrum frequencies using 22 bits without any perceivable spectral distortion.
A significance statement is not available in the OpenAlex record.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
In this contribution, computationally efficient CELP structures are proposed for low-complexity speech coding below 6.4Kbits/s. We also report on a new approach for low bit rate speech coding using binary excitation pulses, which delivers the quality of the CELP codec without the excitation codebook storage, and with a very simple excitation determination procedure. Furthermore, we propose a switched-adaptive stochastic vector quantisation scheme for quantising the line spectrum frequencies using 22 bits without any perceivable spectral distortion.
Key concepts: Code-excited linear prediction, Adaptive Multi-Rate audio codec, Speech coding, Codec2, Vector sum excited linear prediction, Linear predictive coding, Computer science, Harmonic Vector Excitation Coding