Code-excited linear prediction(CELP): High-quality speech at very low bit rates

By M. Schroeder, B. Atal. 1985

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Speech & Audio

Citations in the corpus

SoundStream: An End-to-End Neural Audio Codec (2021, methodology): “For example, CELP [30] adopts an excitation signal encoded via a vector quantizer codebook.”
Cascaded Cross-Module Residual Learning towards Lightweight End-to-End Speech Coding (2019, background): “There have been various conventional speech coding methodologies, including linear predictive coding (LPC) [4], adaptive encoding [5], and perceptual weighting [6] among other domain specific knowledge about the speech signals, that are used to construct classic codecs, such as AMRWB [7] and OPUS [8”
LPCNET: Improving Neural Speech Synthesis through Linear Prediction (2018, methodology): “This is similar to the effect of analysis-by-synthesis in CELP [17, 18] and greatly reduces the artifacts in the synthesized speech.”
Improving Opus Low Bit Rate Quality with Neural Speech Synthesis (2019, background): “SILK is a linear prediction-based coder that uses noise feedback coding (NFC) [20] rather than regular CELP [5].”
Traffic Morphing: An Efficient Defense Against Statistical Traffic Analysis (2009, methodology): “Whitebox vs Blackbox Morphing The Speex encoder is based on a technique called Code Excited Linear Prediction (CELP) [18], in which the sound in each packet is compressed by searching a codebook of excitation vectors for the codebook entry that best reproduces the input speech.”
Neural Feature Predictor and Discriminative Residual Coding for Low-Bitrate Speech Coding (2022, background): “It has been used in various communication and entertainment applications after standardization [1, 2] or open-sourcing [3].”
Techniques for Harmonic Sinusoidal Coding (1997, background): “74 TABLE 6.1: AVERAGE SNR RESULTS FOR PHASE MODELLING SCHEMES........................................................... 108 TABLE 7.1: BIT ALLOCATION OF QUANTISED CODER ....................................................................................... 112 TABLE 7.2: LINEAR QUANTISER PARAMETERS”
Introduction to Digital Speech Processing (2007, background): “The next major innovation in the history of analysis-by-synthesis systems was called code-excited linear predictive coding or CELP [112].”
Unsupervised Speech Decomposition via Triple Information Bottleneck (2020, background): “In speech generation tasks, disentanglement is the foundation of many applications, such as voice conversion (Chou & Lee, 2019), prosody modifcation (Shechtman & Sorin, 2019), emotional speech synthesis (Pell et al., 2011), and low bit-rate speech encoding (Schroeder & Atal, 1985), to name a few.”
Speex: A Free Codec For Free Speech (2016, methodology): “The CELP algorithm [1] was chosen because of its proven track record from low bitrates (e.”
Quasi-Periodic Parallel WaveGAN Vocoder: A Non-autoregressive Pitch-dependent Dilated Convolution Model for Parametric Speech Generation (2020, methodology): “The main modules of the QP structure are PDCNN components and a cascaded architecture, which are inspired by the pitch filtering and the cascaded recursive structure of the code-excited linear prediction (CELP) codec [29].”