Linear predictive coding

from Wikipedia, the free encyclopedia

Linear Predictive Coding ( LPC ) is a method of audio signal and speech processing. a. used for audio data compression and speech recognition . This creates a simplified model of the human vocal tract with which speech can be synthesized . A speech signal can therefore only be described by the model parameters and excitation patterns. This means that significantly less data is generated than with a PCM presentation. Usually an error signal describing the difference between the model prediction and the original signal is also transmitted.

The model of the vocal tract often consists of a vibration generator to simulate the vocal cords or the larynx , and a simple tube as a representation of the downstream vocal tract or articulation space ( throat , mouth). The vibration generator generates a regular vibration, which can be modulated in volume and pitch . The tube represents the resonance space of the throat and mouth and usually serves as a sufficient representation in the case of simple vowels . In contrast, the representation of nasal sounds would require branches from the tube ( nasal cavity ), which would be mathematically much more complex to represent. Therefore these sound components are partly discarded and partly described with the help of a residual signal.

A well-known example of the use of LPC for audio data compression is CELP , which compresses the residual signal by means of code books . The standardized LPC-10e method works without code books and leads to a lower bit rate at the expense of sound quality.

See also

Web links

Individual evidence

  1. otolith.com ( Memento of the original from June 24, 2011 in the Internet Archive ) Info: The archive link was inserted automatically and has not yet been checked. Please check the original and archive link according to the instructions and then remove this notice. @1@ 2Template: Webachiv / IABot / otolith.com