lpc

README

Linear Predictive Coding

Linear predictive coding is a standard method for modelling signals. This package provides a basic implementation based on the autocorrelation method with filter stability enforcement.

Documentation

Overview

Package lpc provides a linear predictive coding interface.

The LPC modelling algorithm is based on the autocorrelation method with the addition of a numerical tweak to enforce stability of the resulting model.

Package lpc supports modelling, predicting and generation/synthesis.

Package lpc does not yet support line spectral frequencies or conversion to other coefficient representations.

Index

• type State
  • func (s *State) Consume(d float64) float64
  • func (s *State) Predict() float64
  • func (s *State) Produce(r float64) float64
• type T
  • func New(order int) *T
  • func (p *T) Model(d []float64) float64
  • func (p *T) Order() int
  • func (p *T) R0() float64
  • func (p *T) Residue(d []float64)
  • func (p *T) Restore(d []float64)
  • func (p *T) State(seed []float64) *State

Types

type State

type State struct {
	// contains filtered or unexported fields
}

State encapsulates linear prediction state for incremental usage in synthesis and prediction.

func (*State) Consume

func (s *State) Consume(d float64) float64

Consume advances the state one element (d), and returns the residue of the model for d.

func (*State) Predict

func (s *State) Predict() float64

Predict returns the current prediction for the next value.

func (*State) Produce

func (s *State) Produce(r float64) float64

Produce synthesizes the next state from the residue r.

type T

type T struct {
	// contains filtered or unexported fields
}

T holds states for doing linear predictive coding of a given order.

func New

func New(order int) *T

New returns a new linear predictive coder.

func (*T) Model

func (p *T) Model(d []float64) float64

Model causes p to learn coeficients for d, returning the model error.

func (*T) Order

func (p *T) Order() int

Order returns the order of p.

func (*T) R0

func (p *T) R0() float64

R0 returns the zero autocorrelation value, useful for normalizing error.

func (*T) Residue

func (p *T) Residue(d []float64)

Residue applies the model to d and for d[p.Order():] replaces the value of the input with the error (aka residue) of the model.

func (*T) Restore

func (p *T) Restore(d []float64)

Restore restores d if d[p.Order():] is a residue generated from d[:p.Order()].

func (*T) State

func (p *T) State(seed []float64) *State

State returns an lpc state for incrementally predicting and synthesizing values according to the model in p.