Qualcomm code-excited linear prediction (QCELP), also known as Qualcomm PureVoice, is a speech codec developed in 1994 by Qualcomm to increase the speech quality of the IS-96A codec earlier used in CDMA networks. It was later replaced with EVRC to achieve better speech quality with fewer bits. The two versions, QCELP8 and QCELP13, operate at 8 and 13 kilobits per second (Kbit/s) respectively.

In CDMA systems, a QCELP vocoder converts a sound signal into a signal transmissible within a circuit. In wired systems, voice signals are generally sampled at 8 kHz (that is, 8,000 sample values per second) and then encoded by 8-bit quantization for each sample value. Such a system transmits at 64 kbit/s, an expensive rate in a wireless system. A QCELP vocoder with variable rates can reduce the rate enough to fit a wireless system by coding the information more efficiently. In particular, it can change its own coding rates based on the speaker's volume or pitch; a louder or higher-pitched voice requires a higher rate.

References

  • https://web.archive.org/web/20130129041829/http://www.rfdh.com/ez/system/db/lib_jnl/upload/2489/[CSD9904]_QCELP_Vocoders_in_CDMA_Systems_Design.pdf
  • Archived PureVoice encoder and player downloads

Codeexcited linear prediction Semantic Scholar

Codeexcited linear prediction Semantic Scholar

Algebraic codeexcited linear prediction Semantic Scholar

Codeexcited linear prediction Semantic Scholar

Codeexcited linear prediction Semantic Scholar