Linear predictive coding is one of a family of analysis/resynthesis techniques developed in the search for techniques for speech encoding, transmission and reception and is still the most widely used technique for computer-generated speech. LPC attempts to estimate the spectrum of a sound in terms of the filter coefficients which would be needed to synthesise that sound when applied to an 'excitation source', which may be either noise (for unvoiced sounds), or a band-limited pulse wave in which all the harmonics are of equal amplitude. In a synthesis application, both the excitation waveform and the filter coefficients can be modified (in both the frequency and time domains) to create related variants of the original sound. (Source - Richard Dobson (1992). A Dictionary of Electronic and Computer Music Technology. Oxford University Press.)

Le codage prédictif linéaire fait partie des méthodes d'analyse/synthèse développées dans le cadre des techniques d'encodage, de transmission et de réception de la parole. De plus, elle est toujours une des techniques les plus utilisées dans la synthèse numérique de la parole. Le CPL est fondé sur une estimation du spectre d'un son en termes de coefficients de filtre qui seraient nécessaires pour synthétiser ce même son une fois appliqués à une “source d'excitation“. Ce son peut être un bruit (sons non vocaux) ou des impulsions limitées à des bandes de fréquences dans lesquelles les harmoniques sont d'égales amplitudes. Dans une application de synthèse, l'excitation des formes d'onde et les coefficients des filtres peuvent être modifiés (à la fois en fréquence et en temps) afin de créer des variations sur les sons originaux. (Source - Richard Dobson (1992). A Dictionary of Electronic and Computer Music Technology. Oxford University Press.)

La codificazione per predizione lineare (Linear Predictive Coding) fa parte di una famiglia di tecniche per la sintesi/risintesi, sviluppate per la codificazione, la trasmissione e la ricezione del linguaggio umano. E' tuttora una delle tecniche più utilizzate per la simulazione al computer del linguaggio. La LPC si basa sulla stima dello spettro di un suono, tramite la definizione dei coefficienti di filtri che sarebbero necessari per sintetizzarlo se applicato ad una fonte di eccitazione. Tale suono può essere sia un rumore, nel caso di suoni non vocali, sia un impulso con bande di frequenze limitate le cui armoniche sono alla stessa ampiezza. In un processo di sintesi con LPC, sia l'onda di eccitazione sia i coefficienti di filtro possono essere modificati (sia nel dominio della frequenza che del tempo), per creare varianti del suono originale (Fonte - Richard Dobson (1992). A Dictionary of Electronic and Computer Music Technology. Oxford University Press).

Linear predictive coding is one of a family of analysis/resynthesis techniques developed in the search for techniques for speech encoding, transmission and reception and is still the most widely used technique for computer-generated speech. LPC attempts to estimate the spectrum of a sound in terms of the filter coefficients which would be needed to synthesise that sound when applied to an 'excitation source', which may be either noise (for unvoiced sounds), or a band-limited pulse wave in which all the harmonics are of equal amplitude. In a synthesis application, both the excitation waveform and the filter coefficients can be modified (in both the frequency and time domains) to create related variants of the original sound. (Source - Richard Dobson (1992). A Dictionary of Electronic and Computer Music Technology. Oxford University Press.)

La síntesis mediante predicción lineal (Linear Predictive Coding o LPC) es una de las técnicas, dentro de la familia de análisis/resíntesis, desarrollada en la búsqueda de métodos de codificación, transmisión y recepción del habla. Es todavía la técnica más ampliamente utilizada en las voces generadas por computadora. La LPC intenta estimar el espectro de un sonido en términos de los coeficientes del filtro que podrían necesitarse para sintetizarlo a partir de una fuente de excitación, ya sea ruido (para los sonidos producido sin vibración de las cuerdas vocales o unvoiced), o un pulso de banda limitada (en la que todos los armónicos son de igual amplitud). Cuando se emplea esta técnica para sintetizar sonidos, tanto la forma de onda usada para la excitación como los coeficientes de filtrado pueden ser modificados (tanto en el dominio temporal como en el de las frecuencias) para crear variaciones del sonido original.

(Fuente: Richard Dobson (1992). A Dictionary of Electronic and Computer Music Technology. Oxford University Press)