DSP Innovations has developed a new technology, called Sinusoidal Pulsed Representation (SPR) for speech coding for low bit rates, from 4,800 bits per second down to 600 bits per second and even lower.

This new technology is based on an advanced parametric model of the speech synthesis where speech is formed as a mixture of fundamental frequency harmonics, the noise-like signal, and aperiodic impulses.

Unlike well-known speech models, such as MELPe, where speech is presented as a mixture of two (voiced and unvoiced) components, the SPR model includes a third component like used in popular Multi-Pulsed Excitation model. This representation allows it to synthesize correctly all speech sounds, including complex speech transients with aperiodic vibration of vocal cords. All parameters of the new model are found on analysis by synthesis base, providing a very high-quality of recovered speech.

SPR technology is available from DSP Innovations in two formats: SPR Classic and SPR Robust. Both are proprietary, royalty-free vocoders. SPR Classic vocoders provide a new level of speech quality, which was unachievable earlier for such low bit rates. These vocoders are suitable ideally for voice storage or voice communications through errorless channels. Unlike SPR Classic vocoders, SPR Robust vocoders include Forward Error Correction (FEC) scheme that is integrated into vocoders using the source-channel coding method and provides very high robustness of the vocoders to channel errors. These vocoders are suitable for voice communications through noisy channels, including radio channels with a very low signal-to-noise ratio. 

About DSP Innovations

DSP Innovations, Inc. (www.dspini.com) was founded in 2007, having united the best Russian scientists, software engineers, and hardware designers who have great experience in DSP and the development of various telecommunication equipment. The mainstream is speech technologies and voice communication: innovative methods of speech coding, speech enhancement (noise and echo cancelling, spectrum equalization, etc.), advanced methods of "source-channel coding". The company is a supplier of DSP-software and engineering services. Proprietary, royalty-free, and standard vocoders from DSPINI have superior characteristics, operate in a range from 300 bps up to 64 kbps and are used in: secure voice, software-defined radio, wireless, VoIP, voice storage, etc. Having great experience in the development of math-intensive algorithms and software, in hardware design, production, testing, and exploitation of various radio and telecommunication equipment, DSPINI's team uses a strong background to provide to each customer the most advantageous solution.