SPR Vocoders: Classic/Robust

SPR Classic™

It is well known, Speech Coding is related to "Source Coding" technology, when redundancy is removed from information source (from speech in this case) to compress its volume and for example to transmit speech through telecommunication channel with limited capacity.

We have named such vocoders as "Classic" and mean these vocoders doesn't contain any redundancy added from outside unlike "Robust" vocoders.

We have developed a new thecnology of speech coding, which have named as Sinusoidal Pulsed Representation (SPR).

The innovative method uses more complex speech model, where excitation of classic LPC-filter includes three components:
- harmonics of the fundamental frequency,
- colored noise,
- aperiodic pulses

Such representation allows to synthesize all speech sounds, including voiced, unvoiced, mixed and also complex transients with aperiodic vibration of vocal cords.

Our proprietary SPR Classic™ vocoders are based on the SPR model and outperform significantly by all characteristics any other similar products known on the market today.

Our "classic series" includes vocoders for 2400, 1200, 1000, 800, 600 and even 300 bps rate.

We are ready to customize any of our vocoder under your specific requirements and conditions to provide the best solution for you.

Classic/Robust Speech Quality

SPR Robust™

It is well known, so-called "channel coding" technology is used to increase significantly robustness of transmission of information through noisy channel. As rule, this coding is boiled down to addition of controlled redundancy to transmitted information. This technology is well-known yet as FEC (Forward Error Correction).

To provide robust transmission of digital speech through noisy channel with limited capacity, in the beginning, speech is compressed by significant removing of uncontrolled redundancy (it is so-called "source coding") and then controlled redundancy of the FEC is added to compressed speech.

However, in majority of cases significant removing of redundancy in the vocoder decreases initial speech quality also significantly. Therefore, such approach as sequential "source coding" and then "channel coding" provides high robustness at the expense of significant lowering of speech quality.

In contrast to it, so-called "source-channel coding" approach solves this tradeoff problem between lowering of initial speech quality and robustness. In this case, vocoder and FEC are developed jointly and redundancy is distributed between vocoder and FEC as "residual" and "controlled" optimally, where criteria of optimality is high speech quality in both clear channel and noisy channel.

We have developed series of SPR Robust™ Vocoders for 800 , 1000, 1200, 2400 and 3600 bps rates as integration of our SPR Classic Vocoders with FEC, using "source-channel coding" approach with minimal lowering of the initial speech quality and have provided therefore as much as possible speech quality in any conditions of communication channel.

We use unregular protection in accordance with informativity=sensitivity of bits in vocoder's bitstream and encode these bits, using Systematic Recursive Convolutional Code (SRCC) with puncturing.

Our robust vocoders are capable to operate with both a "hard decision" and a "soft decision" of information from channel (from demodulator). "Soft decision" mode is preferable because it provide robustness and quality much more in comparison with "hard decision".

In addition, we use "residual redundancy" of vocoder to minimize aftereffects of uncorrectable channel errors.

We are ready to customize our robust vocoders and other technologies for your specific requirements and conditions.