libavcodec/acelp_filters.c File Reference

#include <inttypes.h>
#include "avcodec.h"
#include "acelp_filters.h"

Go to the source code of this file.

Functions
void	ff_acelp_interpolate (int16_t *out, const int16_t *in, const int16_t *filter_coeffs, int precision, int frac_pos, int filter_length, int length)
	Generic FIR interpolation routine.
void	ff_acelp_interpolatef (float *out, const float *in, const float *filter_coeffs, int precision, int frac_pos, int filter_length, int length)
	Floating point version of ff_acelp_interpolate().
void	ff_acelp_high_pass_filter (int16_t *out, int hpf_f[2], const int16_t *in, int length)
	high-pass filtering and upscaling (4.2.5 of G.729).
void	ff_acelp_apply_order_2_transfer_function (float *out, const float *in, const float zero_coeffs[2], const float pole_coeffs[2], float gain, float mem[2], int n)
	Apply an order 2 rational transfer function in-place.
void	ff_tilt_compensation (float *mem, float tilt, float *samples, int size)
	Apply tilt compensation filter, 1 - tilt * z-1.
Variables
const int16_t	ff_acelp_interp_filter [61]
	low-pass Finite Impulse Response filter coefficients.

---

Function Documentation

void ff_acelp_apply_order_2_transfer_function	(	float *	out,
		const float *	in,
		const float	zero_coeffs[2],
		const float	pole_coeffs[2],
		float	gain,
		float	mem[2],
		int	n
	)

Apply an order 2 rational transfer function in-place.

Parameters:

	out	output buffer for filtered speech samples
	in	input buffer containing speech data (may be the same as out)
	zero_coeffs	z^-1 and z^-2 coefficients of the numerator
	pole_coeffs	z^-1 and z^-2 coefficients of the denominator
	gain	scale factor for final output
	mem	intermediate values used by filter (should be 0 initially)
	n	number of samples

Definition at line 117 of file acelp_filters.c.

Referenced by amrnb_decode_frame(), amrwb_decode_frame(), decode_frame(), and postfilter().

void ff_acelp_high_pass_filter	(	int16_t *	out,
		int	hpf_f[2],
		const int16_t *	in,
		int	length
	)

high-pass filtering and upscaling (4.2.5 of G.729).

Parameters:

[out]	out	output buffer for filtered speech data
[in,out]	hpf_f	past filtered data from previous (2 items long) frames (-0x20000000 <= (14.13) < 0x20000000)
	in	speech data to process
	length	input data size

out[i] = 0.93980581 * in[i] - 1.8795834 * in[i-1] + 0.93980581 * in[i-2] + 1.9330735 * out[i-1] - 0.93589199 * out[i-2]

The filter has a cut-off frequency of 1/80 of the sampling freq

Note:

Two items before the top of the out buffer must contain two items from the tail of the previous subframe.

Remarks:

It is safe to pass the same array in in and out parameters.

AMR uses mostly the same filter (cut-off frequency 60Hz, same formula, but constants differs in 5th sign after comma). Fortunately in fixed-point all coefficients are the same as in G.729. Thus this routine can be used for the fixed-point AMR decoder, too.

Definition at line 97 of file acelp_filters.c.

void ff_acelp_interpolate	(	int16_t *	out,
		const int16_t *	in,
		const int16_t *	filter_coeffs,
		int	precision,
		int	frac_pos,
		int	filter_length,
		int	length
	)

Generic FIR interpolation routine.

Parameters:

[out]	out	buffer for interpolated data
	in	input data
	filter_coeffs	interpolation filter coefficients (0.15)
	precision	sub sample factor, that is the precision of the position
	frac_pos	fractional part of position [0..precision-1]
	filter_length	filter length
	length	length of output

filter_coeffs contains coefficients of the right half of the symmetric interpolation filter. filter_coeffs[0] should the central (unpaired) coefficient. See ff_acelp_interp_filter for an example.

Definition at line 42 of file acelp_filters.c.

void ff_acelp_interpolatef	(	float *	out,
		const float *	in,
		const float *	filter_coeffs,
		int	precision,
		int	frac_pos,
		int	filter_length,
		int	length
	)

Floating point version of ff_acelp_interpolate().

Definition at line 76 of file acelp_filters.c.

Referenced by decode_frame(), decode_pitch_vector(), ff_sipr_decode_frame_16k(), and synth_block_fcb_acb().

void ff_tilt_compensation	(	float *	mem,
		float	tilt,
		float *	samples,
		int	size
	)

Apply tilt compensation filter, 1 - tilt * z-1.

Parameters:

	mem	pointer to the filter's state (one single float)
	tilt	tilt factor
	samples	array where the filter is applied
	size	the size of the samples array

Definition at line 134 of file acelp_filters.c.

Referenced by calc_input_response(), postfilter(), postfilter_5k0(), and wiener_denoise().

---

Variable Documentation

const int16_t ff_acelp_interp_filter[61]

Initial value:

 { 
  29443, 28346, 25207, 20449, 14701,  8693,
   3143, -1352, -4402, -5865, -5850, -4673,
  -2783,  -672,  1211,  2536,  3130,  2991,
   2259,  1170,     0, -1001, -1652, -1868,
  -1666, -1147,  -464,   218,   756,  1060,
   1099,   904,   550,   135,  -245,  -514,
   -634,  -602,  -451,  -231,     0,   191,
    308,   340,   296,   198,    78,   -36,
   -120,  -163,  -165,  -132,   -79,   -19,
     34,    73,    91,    89,    70,    38,
      0,
}

low-pass Finite Impulse Response filter coefficients.

Hamming windowed sinc filter with cutoff freq 3/40 of the sampling freq, the coefficients are scaled by 2^15. This array only contains the right half of the filter. This filter is likely identical to the one used in G.729, though this could not be determined from the original comments with certainity.

Definition at line 28 of file acelp_filters.c.