easylinux/OpenTXW81X
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

pure sdk for main

Browse Source
This commit is contained in:
divadiow
2025-08-27 09:51:58 +01:00
parent f0d033f1c9
commit 0571416e7c
3283 changed files with 1577720 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files
sdk/lib/webrtc/modules/audio_processing/aecm/aecm_core.c
+1079
View File
File diff suppressed because it is too large Load Diff
sdk/lib/webrtc/modules/audio_processing/aecm/aecm_core.h
+402
View File
@@ -0,0 +1,402 @@
/*
* Copyright (c) 2012 The WebRTC project authors. All Rights Reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
// Performs echo control (suppression) with fft routines in fixed-point.
#ifndef MODULES_AUDIO_PROCESSING_AECM_AECM_CORE_H_
#define MODULES_AUDIO_PROCESSING_AECM_AECM_CORE_H_
#include "rtc_base/checks.h"
#include <stdint.h>
#include "ring_buffer.h"
#include "signal_processing_library.h"
#include "aecm_defines.h"
struct RealFFT;
#define ALIGN8_BEG
#define ALIGN8_END __attribute__((aligned(8)))
typedef struct {
int16_t real;
int16_t imag;
} ComplexInt16;
typedef struct {
int farBufWritePos;
int farBufReadPos;
int knownDelay;
int lastKnownDelay;
int firstVAD; // Parameter to control poorly initialized channels
RingBuffer *farFrameBuf;
RingBuffer *nearNoisyFrameBuf;
RingBuffer *nearCleanFrameBuf;
RingBuffer *outFrameBuf;
int16_t farBuf[FAR_BUF_LEN];
int16_t mult;
uint32_t seed;
// Delay estimation variables
void *delay_estimator_farend;
void *delay_estimator;
uint16_t currentDelay;
// Far end history variables
// TODO(bjornv): Replace |far_history| with ring_buffer.
uint16_t far_history[PART_LEN1 * MAX_DELAY];
int far_history_pos;
int far_q_domains[MAX_DELAY];
int16_t nlpFlag;
int16_t fixedDelay;
uint32_t totCount;
int16_t dfaCleanQDomain;
int16_t dfaCleanQDomainOld;
int16_t dfaNoisyQDomain;
int16_t dfaNoisyQDomainOld;
int16_t nearLogEnergy[MAX_BUF_LEN];
int16_t farLogEnergy;
int16_t echoAdaptLogEnergy[MAX_BUF_LEN];
int16_t echoStoredLogEnergy[MAX_BUF_LEN];
// The extra 16 or 32 bytes in the following buffers are for alignment based
// Neon code.
// It's designed this way since the current GCC compiler can't align a
// buffer in 16 or 32 byte boundaries properly.
int16_t channelStored_buf[PART_LEN1 + 8];
int16_t channelAdapt16_buf[PART_LEN1 + 8];
int32_t channelAdapt32_buf[PART_LEN1 + 8];
int16_t xBuf_buf[PART_LEN2 + 16]; // farend
int16_t dBufClean_buf[PART_LEN2 + 16]; // nearend
int16_t dBufNoisy_buf[PART_LEN2 + 16]; // nearend
int16_t outBuf_buf[PART_LEN + 8];
// Pointers to the above buffers
int16_t *channelStored;
int16_t *channelAdapt16;
int32_t *channelAdapt32;
int16_t *xBuf;
int16_t *dBufClean;
int16_t *dBufNoisy;
int16_t *outBuf;
int32_t echoFilt[PART_LEN1];
int16_t nearFilt[PART_LEN1];
int32_t noiseEst[PART_LEN1];
int noiseEstTooLowCtr[PART_LEN1];
int noiseEstTooHighCtr[PART_LEN1];
int16_t noiseEstCtr;
int16_t cngMode;
int32_t mseAdaptOld;
int32_t mseStoredOld;
int32_t mseThreshold;
int16_t farEnergyMin;
int16_t farEnergyMax;
int16_t farEnergyMaxMin;
int16_t farEnergyVAD;
int16_t farEnergyMSE;
int currentVADValue;
int16_t vadUpdateCount;
int16_t startupState;
int16_t mseChannelCount;
int16_t supGain;
int16_t supGainOld;
int16_t supGainErrParamA;
int16_t supGainErrParamD;
int16_t supGainErrParamDiffAB;
int16_t supGainErrParamDiffBD;
struct RealFFT *real_fft;
#ifdef AEC_DEBUG
FILE* farFile;
FILE* nearFile;
FILE* outFile;
#endif
} AecmCore;
////////////////////////////////////////////////////////////////////////////////
// WebRtcAecm_CreateCore()
//
// Allocates the memory needed by the AECM. The memory needs to be
// initialized separately using the WebRtcAecm_InitCore() function.
// Returns a pointer to the instance and a nullptr at failure.
AecmCore *WebRtcAecm_CreateCore();
////////////////////////////////////////////////////////////////////////////////
// WebRtcAecm_InitCore(...)
//
// This function initializes the AECM instant created with
// WebRtcAecm_CreateCore()
// Input:
// - aecm : Pointer to the AECM instance
// - samplingFreq : Sampling Frequency
//
// Output:
// - aecm : Initialized instance
//
// Return value : 0 - Ok
// -1 - Error
//
int WebRtcAecm_InitCore(AecmCore *const aecm, int samplingFreq);
////////////////////////////////////////////////////////////////////////////////
// WebRtcAecm_FreeCore(...)
//
// This function releases the memory allocated by WebRtcAecm_CreateCore()
// Input:
// - aecm : Pointer to the AECM instance
//
void WebRtcAecm_FreeCore(AecmCore *aecm);
int WebRtcAecm_Control(AecmCore *aecm, int delay, int nlpFlag);
////////////////////////////////////////////////////////////////////////////////
// WebRtcAecm_InitEchoPathCore(...)
//
// This function resets the echo channel adaptation with the specified channel.
// Input:
// - aecm : Pointer to the AECM instance
// - echo_path : Pointer to the data that should initialize the echo
// path
//
// Output:
// - aecm : Initialized instance
//
void WebRtcAecm_InitEchoPathCore(AecmCore *aecm, const int16_t *echo_path);
////////////////////////////////////////////////////////////////////////////////
// WebRtcAecm_ProcessFrame(...)
//
// This function processes frames and sends blocks to
// WebRtcAecm_ProcessBlock(...)
//
// Inputs:
// - aecm : Pointer to the AECM instance
// - farend : In buffer containing one frame of echo signal
// - nearendNoisy : In buffer containing one frame of nearend+echo signal
// without NS
// - nearendClean : In buffer containing one frame of nearend+echo signal
// with NS
//
// Output:
// - out : Out buffer, one frame of nearend signal :
//
//
int WebRtcAecm_ProcessFrame(AecmCore *aecm,
const int16_t *farend,
const int16_t *nearendNoisy,
const int16_t *nearendClean,
int16_t *out);
////////////////////////////////////////////////////////////////////////////////
// WebRtcAecm_ProcessBlock(...)
//
// This function is called for every block within one frame
// This function is called by WebRtcAecm_ProcessFrame(...)
//
// Inputs:
// - aecm : Pointer to the AECM instance
// - farend : In buffer containing one block of echo signal
// - nearendNoisy : In buffer containing one frame of nearend+echo signal
// without NS
// - nearendClean : In buffer containing one frame of nearend+echo signal
// with NS
//
// Output:
// - out : Out buffer, one block of nearend signal :
//
//
int WebRtcAecm_ProcessBlock(AecmCore *aecm,
const int16_t *farend,
const int16_t *nearendNoisy,
const int16_t *noisyClean,
int16_t *out);
////////////////////////////////////////////////////////////////////////////////
// WebRtcAecm_BufferFarFrame()
//
// Inserts a frame of data into farend buffer.
//
// Inputs:
// - aecm : Pointer to the AECM instance
// - farend : In buffer containing one frame of farend signal
// - farLen : Length of frame
//
void WebRtcAecm_BufferFarFrame(AecmCore *const aecm,
const int16_t *const farend,
const int farLen);
////////////////////////////////////////////////////////////////////////////////
// WebRtcAecm_FetchFarFrame()
//
// Read the farend buffer to account for known delay
//
// Inputs:
// - aecm : Pointer to the AECM instance
// - farend : In buffer containing one frame of farend signal
// - farLen : Length of frame
// - knownDelay : known delay
//
void WebRtcAecm_FetchFarFrame(AecmCore *const aecm,
int16_t *const farend,
const int farLen,
const int knownDelay);
// All the functions below are intended to be private
////////////////////////////////////////////////////////////////////////////////
// WebRtcAecm_UpdateFarHistory()
//
// Moves the pointer to the next entry and inserts |far_spectrum| and
// corresponding Q-domain in its buffer.
//
// Inputs:
// - self : Pointer to the delay estimation instance
// - far_spectrum : Pointer to the far end spectrum
// - far_q : Q-domain of far end spectrum
//
void WebRtcAecm_UpdateFarHistory(AecmCore *self,
uint16_t *far_spectrum,
int far_q);
////////////////////////////////////////////////////////////////////////////////
// WebRtcAecm_AlignedFarend()
//
// Returns a pointer to the far end spectrum aligned to current near end
// spectrum. The function WebRtc_DelayEstimatorProcessFix(...) should have been
// called before AlignedFarend(...). Otherwise, you get the pointer to the
// previous frame. The memory is only valid until the next call of
// WebRtc_DelayEstimatorProcessFix(...).
//
// Inputs:
// - self : Pointer to the AECM instance.
// - delay : Current delay estimate.
//
// Output:
// - far_q : The Q-domain of the aligned far end spectrum
//
// Return value:
// - far_spectrum : Pointer to the aligned far end spectrum
// NULL - Error
//
const uint16_t *WebRtcAecm_AlignedFarend(AecmCore *self, int *far_q, int delay);
///////////////////////////////////////////////////////////////////////////////
// WebRtcAecm_CalcSuppressionGain()
//
// This function calculates the suppression gain that is used in the
// Wiener filter.
//
// Inputs:
// - aecm : Pointer to the AECM instance.
//
// Return value:
// - supGain : Suppression gain with which to scale the noise
// level (Q14).
//
int16_t WebRtcAecm_CalcSuppressionGain(AecmCore *const aecm);
///////////////////////////////////////////////////////////////////////////////
// WebRtcAecm_CalcEnergies()
//
// This function calculates the log of energies for nearend, farend and
// estimated echoes. There is also an update of energy decision levels,
// i.e. internal VAD.
//
// Inputs:
// - aecm : Pointer to the AECM instance.
// - far_spectrum : Pointer to farend spectrum.
// - far_q : Q-domain of farend spectrum.
// - nearEner : Near end energy for current block in
// Q(aecm->dfaQDomain).
//
// Output:
// - echoEst : Estimated echo in Q(xfa_q+RESOLUTION_CHANNEL16).
//
void WebRtcAecm_CalcEnergies(AecmCore *aecm,
const uint16_t *far_spectrum,
const int16_t far_q,
const uint32_t nearEner,
int32_t *echoEst);
///////////////////////////////////////////////////////////////////////////////
// WebRtcAecm_CalcStepSize()
//
// This function calculates the step size used in channel estimation
//
// Inputs:
// - aecm : Pointer to the AECM instance.
//
// Return value:
// - mu : Stepsize in log2(), i.e. number of shifts.
//
int16_t WebRtcAecm_CalcStepSize(AecmCore *const aecm);
///////////////////////////////////////////////////////////////////////////////
// WebRtcAecm_UpdateChannel(...)
//
// This function performs channel estimation.
// NLMS and decision on channel storage.
//
// Inputs:
// - aecm : Pointer to the AECM instance.
// - far_spectrum : Absolute value of the farend signal in Q(far_q)
// - far_q : Q-domain of the farend signal
// - dfa : Absolute value of the nearend signal
// (Q[aecm->dfaQDomain])
// - mu : NLMS step size.
// Input/Output:
// - echoEst : Estimated echo in Q(far_q+RESOLUTION_CHANNEL16).
//
void WebRtcAecm_UpdateChannel(AecmCore *aecm,
const uint16_t *far_spectrum,
const int16_t far_q,
const uint16_t *const dfa,
const int16_t mu,
int32_t *echoEst);
extern const int16_t WebRtcAecm_kCosTable[];
extern const int16_t WebRtcAecm_kSinTable[];
///////////////////////////////////////////////////////////////////////////////
// Some function pointers, for internal functions shared by ARM NEON and
// generic C code.
//
typedef void (*CalcLinearEnergies)(AecmCore *aecm,
const uint16_t *far_spectrum,
int32_t *echoEst,
uint32_t *far_energy,
uint32_t *echo_energy_adapt,
uint32_t *echo_energy_stored);
extern CalcLinearEnergies WebRtcAecm_CalcLinearEnergies;
typedef void (*StoreAdaptiveChannel)(AecmCore *aecm,
const uint16_t *far_spectrum,
int32_t *echo_est);
extern StoreAdaptiveChannel WebRtcAecm_StoreAdaptiveChannel;
typedef void (*ResetAdaptiveChannel)(AecmCore *aecm);
extern ResetAdaptiveChannel WebRtcAecm_ResetAdaptiveChannel;
#endif
sdk/lib/webrtc/modules/audio_processing/aecm/aecm_core_c.c
+662
View File
@@ -0,0 +1,662 @@
/*
* Copyright (c) 2013 The WebRTC project authors. All Rights Reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
#include <stddef.h>
#include "aecm_core.h"
#include "real_fft.h"
#include "echo_control_mobile.h"
#include "delay_estimator_wrapper.h"
// Square root of Hanning window in Q14.
static const ALIGN8_BEG int16_t
WebRtcAecm_kSqrtHanning[]
ALIGN8_END = {
0, 399, 798, 1196, 1594, 1990, 2386, 2780, 3172, 3562, 3951,
4337, 4720, 5101, 5478, 5853, 6224, 6591, 6954, 7313, 7668, 8019,
8364, 8705, 9040, 9370, 9695, 10013, 10326, 10633, 10933, 11227, 11514,
11795, 12068, 12335, 12594, 12845, 13089, 13325, 13553, 13773, 13985, 14189,
14384, 14571, 14749, 14918, 15079, 15231, 15373, 15506, 15631, 15746, 15851,
15947, 16034, 16111, 16179, 16237, 16286, 16325, 16354, 16373, 16384};
#ifdef AECM_WITH_ABS_APPROX
// Q15 alpha = 0.99439986968132 const Factor for magnitude approximation
static const uint16_t kAlpha1 = 32584;
// Q15 beta = 0.12967166976970 const Factor for magnitude approximation
static const uint16_t kBeta1 = 4249;
// Q15 alpha = 0.94234827210087 const Factor for magnitude approximation
static const uint16_t kAlpha2 = 30879;
// Q15 beta = 0.33787806009150 const Factor for magnitude approximation
static const uint16_t kBeta2 = 11072;
// Q15 alpha = 0.82247698684306 const Factor for magnitude approximation
static const uint16_t kAlpha3 = 26951;
// Q15 beta = 0.57762063060713 const Factor for magnitude approximation
static const uint16_t kBeta3 = 18927;
#endif
static const int16_t kNoiseEstQDomain = 15;
static const int16_t kNoiseEstIncCount = 5;
static void ComfortNoise(AecmCore *aecm,
const uint16_t *dfa,
ComplexInt16 *out,
const int16_t *lambda) {
int16_t i;
int16_t tmp16;
int32_t tmp32;
int16_t randW16[PART_LEN];
int16_t uReal[PART_LEN1];
int16_t uImag[PART_LEN1];
int32_t outLShift32;
int16_t noiseRShift16[PART_LEN1];
int16_t shiftFromNearToNoise = kNoiseEstQDomain - aecm->dfaCleanQDomain;
int16_t minTrackShift;
RTC_DCHECK_GE(shiftFromNearToNoise, 0);
RTC_DCHECK_LT(shiftFromNearToNoise, 16);
if (aecm->noiseEstCtr < 100) {
// Track the minimum more quickly initially.
aecm->noiseEstCtr++;
minTrackShift = 6;
} else {
minTrackShift = 9;
}
// Estimate noise power.
for (i = 0; i < PART_LEN1; i++) {
// Shift to the noise domain.
tmp32 = (int32_t) dfa[i];
outLShift32 = tmp32 << shiftFromNearToNoise;
if (outLShift32 < aecm->noiseEst[i]) {
// Reset "too low" counter
aecm->noiseEstTooLowCtr[i] = 0;
// Track the minimum.
if (aecm->noiseEst[i] < (1 << minTrackShift)) {
// For small values, decrease noiseEst[i] every
// |kNoiseEstIncCount| block. The regular approach below can not
// go further down due to truncation.
aecm->noiseEstTooHighCtr[i]++;
if (aecm->noiseEstTooHighCtr[i] >= kNoiseEstIncCount) {
aecm->noiseEst[i]--;
aecm->noiseEstTooHighCtr[i] = 0; // Reset the counter
}
} else {
aecm->noiseEst[i] -=
((aecm->noiseEst[i] - outLShift32) >> minTrackShift);
}
} else {
// Reset "too high" counter
aecm->noiseEstTooHighCtr[i] = 0;
// Ramp slowly upwards until we hit the minimum again.
if ((aecm->noiseEst[i] >> 19) > 0) {
// Avoid overflow.
// Multiplication with 2049 will cause wrap around. Scale
// down first and then multiply
aecm->noiseEst[i] >>= 11;
aecm->noiseEst[i] *= 2049;
} else if ((aecm->noiseEst[i] >> 11) > 0) {
// Large enough for relative increase
aecm->noiseEst[i] *= 2049;
aecm->noiseEst[i] >>= 11;
} else {
// Make incremental increases based on size every
// |kNoiseEstIncCount| block
aecm->noiseEstTooLowCtr[i]++;
if (aecm->noiseEstTooLowCtr[i] >= kNoiseEstIncCount) {
aecm->noiseEst[i] += (aecm->noiseEst[i] >> 9) + 1;
aecm->noiseEstTooLowCtr[i] = 0; // Reset counter
}
}
}
}
for (i = 0; i < PART_LEN1; i++) {
tmp32 = aecm->noiseEst[i] >> shiftFromNearToNoise;
if (tmp32 > 32767) {
tmp32 = 32767;
aecm->noiseEst[i] = tmp32 << shiftFromNearToNoise;
}
noiseRShift16[i] = (int16_t)
tmp32;
tmp16 = ONE_Q14 - lambda[i];
noiseRShift16[i] = (int16_t) ((tmp16 * noiseRShift16[i]) >> 14);
}
// Generate a uniform random array on [0 2^15-1].
WebRtcSpl_RandUArray(randW16, PART_LEN, &aecm->seed);
// Generate noise according to estimated energy.
uReal[0] = 0; // Reject LF noise.
uImag[0] = 0;
for (i = 1; i < PART_LEN1; i++) {
// Get a random index for the cos and sin tables over [0 359].
tmp16 = (int16_t) ((359 * randW16[i - 1]) >> 15);
// Tables are in Q13.
uReal[i] =
(int16_t) ((noiseRShift16[i] * WebRtcAecm_kCosTable[tmp16]) >> 13);
uImag[i] =
(int16_t) ((-noiseRShift16[i] * WebRtcAecm_kSinTable[tmp16]) >> 13);
}
uImag[PART_LEN] = 0;
for (i = 0; i < PART_LEN1; i++) {
out[i].real = WebRtcSpl_AddSatW16(out[i].real, uReal[i]);
out[i].imag = WebRtcSpl_AddSatW16(out[i].imag, uImag[i]);
}
}
static void WindowAndFFT(AecmCore *aecm,
int16_t *fft,
const int16_t *time_signal,
ComplexInt16 *freq_signal,
int time_signal_scaling) {
int i = 0;
// FFT of signal
for (i = 0; i < PART_LEN; i++) {
// Window time domain signal and insert into real part of
// transformation array |fft|
int16_t scaled_time_signal = time_signal[i] * (1 << time_signal_scaling);
fft[i] = (int16_t) ((scaled_time_signal * WebRtcAecm_kSqrtHanning[i]) >> 14);
scaled_time_signal = time_signal[i + PART_LEN] * (1 << time_signal_scaling);
fft[PART_LEN + i] = (int16_t) (
(scaled_time_signal * WebRtcAecm_kSqrtHanning[PART_LEN - i]) >> 14);
}
// Do forward FFT, then take only the first PART_LEN complex samples,
// and change signs of the imaginary parts.
WebRtcSpl_RealForwardFFT(aecm->real_fft, fft, (int16_t *)
freq_signal);
for (i = 0; i < PART_LEN; i++) {
freq_signal[i].imag = -freq_signal[i].imag;
}
}
static void InverseFFTAndWindow(AecmCore *aecm,
int16_t *fft,
ComplexInt16 *efw,
int16_t *output,
const int16_t *nearendClean) {
int i, j, outCFFT;
int32_t tmp32no1;
// Reuse |efw| for the inverse FFT output after transferring
// the contents to |fft|.
int16_t *ifft_out = (int16_t *)
efw;
// Synthesis
for (i = 1, j = 2; i < PART_LEN; i += 1, j += 2) {
fft[j] = efw[i].real;
fft[j + 1] = -efw[i].imag;
}
fft[0] = efw[0].real;
fft[1] = -efw[0].imag;
fft[PART_LEN2] = efw[PART_LEN].real;
fft[PART_LEN2 + 1] = -efw[PART_LEN].imag;
// Inverse FFT. Keep outCFFT to scale the samples in the next block.
outCFFT = WebRtcSpl_RealInverseFFT(aecm->real_fft, fft, ifft_out);
for (i = 0; i < PART_LEN; i++) {
ifft_out[i] = (int16_t)
WEBRTC_SPL_MUL_16_16_RSFT_WITH_ROUND(
ifft_out[i], WebRtcAecm_kSqrtHanning[i], 14);
tmp32no1 = WEBRTC_SPL_SHIFT_W32((int32_t) ifft_out[i],
outCFFT - aecm->dfaCleanQDomain);
output[i] = (int16_t)
WEBRTC_SPL_SAT(WEBRTC_SPL_WORD16_MAX,
tmp32no1 + aecm->outBuf[i],
WEBRTC_SPL_WORD16_MIN);
tmp32no1 =
(ifft_out[PART_LEN + i] * WebRtcAecm_kSqrtHanning[PART_LEN - i]) >> 14;
tmp32no1 = WEBRTC_SPL_SHIFT_W32(tmp32no1, outCFFT - aecm->dfaCleanQDomain);
aecm->outBuf[i] = (int16_t)
WEBRTC_SPL_SAT(WEBRTC_SPL_WORD16_MAX, tmp32no1,
WEBRTC_SPL_WORD16_MIN);
}
// Copy the current block to the old position
// (aecm->outBuf is shifted elsewhere)
memcpy(aecm->xBuf, aecm->xBuf + PART_LEN, sizeof(int16_t) * PART_LEN);
memcpy(aecm->dBufNoisy, aecm->dBufNoisy + PART_LEN,
sizeof(int16_t) * PART_LEN);
if (nearendClean != NULL) {
memcpy(aecm->dBufClean, aecm->dBufClean + PART_LEN,
sizeof(int16_t) * PART_LEN);
}
}
// Transforms a time domain signal into the frequency domain, outputting the
// complex valued signal, absolute value and sum of absolute values.
//
// time_signal [in] Pointer to time domain signal
// freq_signal_real [out] Pointer to real part of frequency domain array
// freq_signal_imag [out] Pointer to imaginary part of frequency domain
// array
// freq_signal_abs [out] Pointer to absolute value of frequency domain
// array
// freq_signal_sum_abs [out] Pointer to the sum of all absolute values in
// the frequency domain array
// return value The Q-domain of current frequency values
//
static int TimeToFrequencyDomain(AecmCore *aecm,
const int16_t *time_signal,
ComplexInt16 *freq_signal,
uint16_t *freq_signal_abs,
uint32_t *freq_signal_sum_abs) {
int i = 0;
int time_signal_scaling = 0;
int32_t tmp32no1 = 0;
int32_t tmp32no2 = 0;
// In fft_buf, +16 for 32-byte alignment.
int16_t fft_buf[PART_LEN4 + 16];
int16_t *fft = (int16_t *) (((uintptr_t) fft_buf + 31) & ~31);
int16_t tmp16no1;
int16_t tmp16no2;
WindowAndFFT(aecm, fft, time_signal, freq_signal, time_signal_scaling);
// Extract imaginary and real part, calculate the magnitude for
// all frequency bins
freq_signal[0].imag = 0;
freq_signal[PART_LEN].imag = 0;
freq_signal_abs[0] = (uint16_t) WEBRTC_SPL_ABS_W16(freq_signal[0].real);
freq_signal_abs[PART_LEN] =
(uint16_t) WEBRTC_SPL_ABS_W16(freq_signal[PART_LEN].real);
(*freq_signal_sum_abs) =
(uint32_t) (freq_signal_abs[0]) + (uint32_t) (freq_signal_abs[PART_LEN]);
for (i = 1; i < PART_LEN; i++) {
if (freq_signal[i].real == 0) {
freq_signal_abs[i] = (uint16_t) WEBRTC_SPL_ABS_W16(freq_signal[i].imag);
} else if (freq_signal[i].imag == 0) {
freq_signal_abs[i] = (uint16_t) WEBRTC_SPL_ABS_W16(freq_signal[i].real);
} else {
// Approximation for magnitude of complex fft output
// magn = sqrt(real^2 + imag^2)
// magn ~= alpha * max(|imag|,|real|) + beta * min(|imag|,|real|)
//
// The parameters alpha and beta are stored in Q15
tmp16no1 = WEBRTC_SPL_ABS_W16(freq_signal[i].real);
tmp16no2 = WEBRTC_SPL_ABS_W16(freq_signal[i].imag);
tmp32no1 = tmp16no1 * tmp16no1;
tmp32no2 = tmp16no2 * tmp16no2;
tmp32no2 = WebRtcSpl_AddSatW32(tmp32no1, tmp32no2);
tmp32no1 = WebRtcSpl_SqrtFloor(tmp32no2);
freq_signal_abs[i] = (uint16_t) tmp32no1;
}
(*freq_signal_sum_abs) += (uint32_t) freq_signal_abs[i];
}
return time_signal_scaling;
}
// bugs.webrtc.org/8200
int WebRtcAecm_ProcessBlock(AecmCore *aecm, const int16_t *farend, const int16_t *nearendNoisy,
const int16_t *nearendClean, int16_t *output) {
int i;
uint32_t xfaSum;
uint32_t dfaNoisySum;
uint32_t dfaCleanSum;
uint32_t echoEst32Gained;
uint32_t tmpU32;
int32_t tmp32no1;
uint16_t xfa[PART_LEN1];
uint16_t dfaNoisy[PART_LEN1];
uint16_t dfaClean[PART_LEN1];
uint16_t *ptrDfaClean = dfaClean;
const uint16_t *far_spectrum_ptr = NULL;
// 32 byte aligned buffers (with +8 or +16).
// TODO(kma): define fft with ComplexInt16.
int16_t fft_buf[PART_LEN4 + 2 + 16]; // +2 to make a loop safe.
int32_t echoEst32_buf[PART_LEN1 + 8];
int32_t dfw_buf[PART_LEN2 + 8];
int32_t efw_buf[PART_LEN2 + 8];
int16_t *fft = (int16_t *) (((uintptr_t) fft_buf + 31) & ~31);
int32_t *echoEst32 = (int32_t *) (((uintptr_t) echoEst32_buf + 31) & ~31);
ComplexInt16 *dfw = (ComplexInt16 *) (((uintptr_t) dfw_buf + 31) & ~31);
ComplexInt16 *efw = (ComplexInt16 *) (((uintptr_t) efw_buf + 31) & ~31);
int16_t hnl[PART_LEN1];
int16_t numPosCoef = 0;
int16_t nlpGain = ONE_Q14;
int delay;
int16_t tmp16no1;
int16_t tmp16no2;
int16_t mu;
int16_t supGain;
int16_t zeros32, zeros16;
int16_t zerosDBufNoisy, zerosDBufClean, zerosXBuf;
int far_q;
int16_t resolutionDiff, qDomainDiff, dfa_clean_q_domain_diff;
const int kMinPrefBand = 4;
const int kMaxPrefBand = 24;
int32_t avgHnl32 = 0;
// Determine startup state. There are three states:
// (0) the first CONV_LEN blocks
// (1) another CONV_LEN blocks
// (2) the rest
if (aecm->startupState < 2) {
aecm->
startupState =
(aecm->totCount >= CONV_LEN) + (aecm->totCount >= CONV_LEN2);
}
// END: Determine startup state
// Buffer near and far end signals
memcpy(aecm
->xBuf + PART_LEN, farend, sizeof(int16_t) * PART_LEN);
memcpy(aecm
->dBufNoisy + PART_LEN, nearendNoisy, sizeof(int16_t) * PART_LEN);
if (nearendClean != NULL) {
memcpy(aecm
->dBufClean + PART_LEN, nearendClean,
sizeof(int16_t) * PART_LEN);
}
// Transform far end signal from time domain to frequency domain.
far_q = TimeToFrequencyDomain(aecm, aecm->xBuf, dfw, xfa, &xfaSum);
// Transform noisy near end signal from time domain to frequency domain.
zerosDBufNoisy =
TimeToFrequencyDomain(aecm, aecm->dBufNoisy, dfw, dfaNoisy, &dfaNoisySum);
aecm->
dfaNoisyQDomainOld = aecm->dfaNoisyQDomain;
aecm->
dfaNoisyQDomain = (int16_t) zerosDBufNoisy;
if (nearendClean == NULL) {
ptrDfaClean = dfaNoisy;
aecm->
dfaCleanQDomainOld = aecm->dfaNoisyQDomainOld;
aecm->
dfaCleanQDomain = aecm->dfaNoisyQDomain;
dfaCleanSum = dfaNoisySum;
} else {
// Transform clean near end signal from time domain to frequency domain.
zerosDBufClean = TimeToFrequencyDomain(aecm, aecm->dBufClean, dfw, dfaClean,
&dfaCleanSum);
aecm->
dfaCleanQDomainOld = aecm->dfaCleanQDomain;
aecm->
dfaCleanQDomain = (int16_t) zerosDBufClean;
}
// Get the delay
// Save far-end history and estimate delay
WebRtcAecm_UpdateFarHistory(aecm, xfa, far_q
);
if (WebRtc_AddFarSpectrumFix(aecm
->delay_estimator_farend, xfa, PART_LEN1,
far_q) == -1) {
return -1;
}
delay = WebRtc_DelayEstimatorProcessFix(aecm->delay_estimator, dfaNoisy,
PART_LEN1, zerosDBufNoisy);
if (delay == -1) {
return -1;
} else if (delay == -2) {
// If the delay is unknown, we assume zero.
// NOTE: this will have to be adjusted if we ever add lookahead.
delay = 0;
}
if (aecm->fixedDelay >= 0) {
// Use fixed delay
delay = aecm->fixedDelay;
}
// Get aligned far end spectrum
far_spectrum_ptr = WebRtcAecm_AlignedFarend(aecm, &far_q, delay);
zerosXBuf = (int16_t) far_q;
if (far_spectrum_ptr == NULL) {
return -1;
}
// Calculate log(energy) and update energy threshold levels
WebRtcAecm_CalcEnergies(aecm, far_spectrum_ptr, zerosXBuf, dfaNoisySum,
echoEst32
);
// Calculate stepsize
mu = WebRtcAecm_CalcStepSize(aecm);
// Update counters
aecm->totCount++;
// This is the channel estimation algorithm.
// It is base on NLMS but has a variable step length,
// which was calculated above.
WebRtcAecm_UpdateChannel(aecm, far_spectrum_ptr, zerosXBuf, dfaNoisy, mu,
echoEst32
);
supGain = WebRtcAecm_CalcSuppressionGain(aecm);
// Calculate Wiener filter hnl[]
for (
i = 0;
i < PART_LEN1;
i++) {
// Far end signal through channel estimate in Q8
// How much can we shift right to preserve resolution
tmp32no1 = echoEst32[i] - aecm->echoFilt[i];
aecm->echoFilt[i] += (int32_t) (((int64_t) tmp32no1* 50) >> 8);
zeros32 = WebRtcSpl_NormW32(aecm->echoFilt[i]) + 1;
zeros16 = WebRtcSpl_NormW16(supGain) + 1;
if (zeros32 + zeros16 > 16) {
// Multiplication is safe
// Result in
// Q(RESOLUTION_CHANNEL+RESOLUTION_SUPGAIN+
// aecm->xfaQDomainBuf[diff])
echoEst32Gained =
WEBRTC_SPL_UMUL_32_16((uint32_t) aecm->echoFilt[i], (uint16_t) supGain);
resolutionDiff = 14 - RESOLUTION_CHANNEL16 - RESOLUTION_SUPGAIN;
resolutionDiff += (aecm->dfaCleanQDomain - zerosXBuf);
} else {
tmp16no1 = 17 - zeros32 - zeros16;
resolutionDiff =
14 + tmp16no1 - RESOLUTION_CHANNEL16 - RESOLUTION_SUPGAIN;
resolutionDiff += (aecm->dfaCleanQDomain - zerosXBuf);
if (zeros32 > tmp16no1) {
echoEst32Gained = WEBRTC_SPL_UMUL_32_16((uint32_t) aecm->echoFilt[i],
supGain >> tmp16no1);
} else {
// Result in Q-(RESOLUTION_CHANNEL+RESOLUTION_SUPGAIN-16)
echoEst32Gained = (aecm->echoFilt[i] >> tmp16no1) * supGain;
}
}
zeros16 = WebRtcSpl_NormW16(aecm->nearFilt[i]);
RTC_DCHECK_GE(zeros16,
0); // |zeros16| is a norm, hence non-negative.
dfa_clean_q_domain_diff = aecm->dfaCleanQDomain - aecm->dfaCleanQDomainOld;
if (zeros16 < dfa_clean_q_domain_diff && aecm->nearFilt[i]) {
tmp16no1 = aecm->nearFilt[i] * (1 << zeros16);
qDomainDiff = zeros16 - dfa_clean_q_domain_diff;
tmp16no2 = ptrDfaClean[i] >> -qDomainDiff;
} else {
tmp16no1 = dfa_clean_q_domain_diff < 0
? aecm->nearFilt[i] >> -dfa_clean_q_domain_diff
: aecm->nearFilt[i] * (1 << dfa_clean_q_domain_diff);
qDomainDiff = 0;
tmp16no2 = ptrDfaClean[i];
}
tmp32no1 = (int32_t) (tmp16no2 - tmp16no1);
tmp16no2 = (int16_t) (tmp32no1 >> 4);
tmp16no2 +=
tmp16no1;
zeros16 = WebRtcSpl_NormW16(tmp16no2);
if ((tmp16no2) & (-qDomainDiff > zeros16)) {
aecm->nearFilt[i] =
WEBRTC_SPL_WORD16_MAX;
} else {
aecm->nearFilt[i] = qDomainDiff < 0 ? tmp16no2 * (1 << -qDomainDiff)
: tmp16no2 >>
qDomainDiff;
}
// Wiener filter coefficients, resulting hnl in Q14
if (echoEst32Gained == 0) {
hnl[i] =
ONE_Q14;
} else if (aecm->nearFilt[i] == 0) {
hnl[i] = 0;
} else {
// Multiply the suppression gain
// Rounding
echoEst32Gained += (uint32_t) (aecm->nearFilt[i] >> 1);
tmpU32 =
WebRtcSpl_DivU32U16(echoEst32Gained, (uint16_t) aecm->nearFilt[i]);
// Current resolution is
// Q-(RESOLUTION_CHANNEL+RESOLUTION_SUPGAIN- max(0,17-zeros16- zeros32))
// Make sure we are in Q14
tmp32no1 = (int32_t) WEBRTC_SPL_SHIFT_W32(tmpU32, resolutionDiff);
if (tmp32no1 > ONE_Q14) {
hnl[i] = 0;
} else if (tmp32no1 < 0) {
hnl[i] =
ONE_Q14;
} else {
// 1-echoEst/dfa
hnl[i] = ONE_Q14 - (int16_t)
tmp32no1;
if (hnl[i] < 0) {
hnl[i] = 0;
}
}
}
if (hnl[i]) {
numPosCoef++;
}
}
// Only in wideband. Prevent the gain in upper band from being larger than
// in lower band.
if (aecm->mult == 2) {
// TODO(bjornv): Investigate if the scaling of hnl[i] below can cause
// speech distortion in double-talk.
for (
i = 0;
i < PART_LEN1;
i++) {
hnl[i] = (int16_t) ((hnl[i] * hnl[i]) >> 14);
}
for (
i = kMinPrefBand;
i <=
kMaxPrefBand;
i++) {
avgHnl32 += (int32_t) hnl[i];
}
RTC_DCHECK_GT(kMaxPrefBand
-kMinPrefBand + 1, 0);
avgHnl32 /= (kMaxPrefBand - kMinPrefBand + 1);
for (
i = kMaxPrefBand;
i < PART_LEN1;
i++) {
if (hnl[i] > (int16_t) avgHnl32) {
hnl[i] = (int16_t)
avgHnl32;
}
}
}
// Calculate NLP gain, result is in Q14
if (aecm->nlpFlag) {
for (
i = 0;
i < PART_LEN1;
i++) {
// Truncate values close to zero and one.
if (hnl[i] > NLP_COMP_HIGH) {
hnl[i] =
ONE_Q14;
} else if (hnl[i] < NLP_COMP_LOW) {
hnl[i] = 0;
}
// Remove outliers
if (numPosCoef < 3) {
nlpGain = 0;
} else {
nlpGain = ONE_Q14;
}
// NLP
if ((hnl[i] == ONE_Q14) && (nlpGain == ONE_Q14)) {
hnl[i] =
ONE_Q14;
} else {
hnl[i] = (int16_t) ((hnl[i] * nlpGain) >> 14);
}
// multiply with Wiener coefficients
efw[i].
real = (int16_t) (
WEBRTC_SPL_MUL_16_16_RSFT_WITH_ROUND(dfw[i].real, hnl[i], 14));
efw[i].
imag = (int16_t) (
WEBRTC_SPL_MUL_16_16_RSFT_WITH_ROUND(dfw[i].imag, hnl[i], 14));
}
} else {
// multiply with Wiener coefficients
for (
i = 0;
i < PART_LEN1;
i++) {
efw[i].
real = (int16_t) (
WEBRTC_SPL_MUL_16_16_RSFT_WITH_ROUND(dfw[i].real, hnl[i], 14));
efw[i].
imag = (int16_t) (
WEBRTC_SPL_MUL_16_16_RSFT_WITH_ROUND(dfw[i].imag, hnl[i], 14));
}
}
if (aecm->cngMode == AecmTrue) {
ComfortNoise(aecm, ptrDfaClean, efw, hnl
);
}
InverseFFTAndWindow(aecm, fft, efw, output, nearendClean
);
return 0;
}
sdk/lib/webrtc/modules/audio_processing/aecm/aecm_defines.h
+87
View File
@@ -0,0 +1,87 @@
/*
* Copyright (c) 2012 The WebRTC project authors. All Rights Reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
#ifndef MODULES_AUDIO_PROCESSING_AECM_AECM_DEFINES_H_
#define MODULES_AUDIO_PROCESSING_AECM_AECM_DEFINES_H_
#define AECM_DYNAMIC_Q /* Turn on/off dynamic Q-domain. */
/* Algorithm parameters */
#define FRAME_LEN 80 /* Total frame length, 10 ms. */
#define PART_LEN 64 /* Length of partition. */
#define PART_LEN_SHIFT 7 /* Length of (PART_LEN * 2) in base 2. */
#define PART_LEN1 (PART_LEN + 1) /* Unique fft coefficients. */
#define PART_LEN2 (PART_LEN << 1) /* Length of partition * 2. */
#define PART_LEN4 (PART_LEN << 2) /* Length of partition * 4. */
#define FAR_BUF_LEN PART_LEN4 /* Length of buffers. */
#define MAX_DELAY 50
/* Counter parameters */
#define CONV_LEN 512 /* Convergence length used at startup. */
#define CONV_LEN2 (CONV_LEN << 1) /* Used at startup. */
/* Energy parameters */
#define MAX_BUF_LEN 64 /* History length of energy signals. */
#define FAR_ENERGY_MIN 1025 /* Lowest Far energy level: At least 2 */
/* in energy. */
#define FAR_ENERGY_DIFF 929 /* Allowed difference between max */
/* and min. */
#define ENERGY_DEV_OFFSET 0 /* The energy error offset in Q8. */
#define ENERGY_DEV_TOL 400 /* The energy estimation tolerance (Q8). */
#define FAR_ENERGY_VAD_REGION 230 /* Far VAD tolerance region. */
/* Stepsize parameters */
#define MU_MIN 10 /* Min stepsize 2^-MU_MIN (far end energy */
/* dependent). */
#define MU_MAX 1 /* Max stepsize 2^-MU_MAX (far end energy */
/* dependent). */
#define MU_DIFF 9 /* MU_MIN - MU_MAX */
/* Channel parameters */
#define MIN_MSE_COUNT 20 /* Min number of consecutive blocks with enough */
/* far end energy to compare channel estimates. */
#define MIN_MSE_DIFF 29 /* The ratio between adapted and stored channel to */
/* accept a new storage (0.8 in Q-MSE_RESOLUTION). */
#define MSE_RESOLUTION 5 /* MSE parameter resolution. */
#define RESOLUTION_CHANNEL16 12 /* W16 Channel in Q-RESOLUTION_CHANNEL16. */
#define RESOLUTION_CHANNEL32 28 /* W32 Channel in Q-RESOLUTION_CHANNEL. */
#define CHANNEL_VAD 16 /* Minimum energy in frequency band */
/* to update channel. */
/* Suppression gain parameters: SUPGAIN parameters in Q-(RESOLUTION_SUPGAIN). */
#define RESOLUTION_SUPGAIN 8 /* Channel in Q-(RESOLUTION_SUPGAIN). */
#define SUPGAIN_DEFAULT (1 << RESOLUTION_SUPGAIN) /* Default. */
#define SUPGAIN_ERROR_PARAM_A 3072 /* Estimation error parameter */
/* (Maximum gain) (8 in Q8). */
#define SUPGAIN_ERROR_PARAM_B 1536 /* Estimation error parameter */
/* (Gain before going down). */
#define SUPGAIN_ERROR_PARAM_D SUPGAIN_DEFAULT /* Estimation error parameter */
/* (Should be the same as Default) (1 in Q8). */
#define SUPGAIN_EPC_DT 200 /* SUPGAIN_ERROR_PARAM_C * ENERGY_DEV_TOL */
/* Defines for "check delay estimation" */
#define CORR_WIDTH 31 /* Number of samples to correlate over. */
#define CORR_MAX 16 /* Maximum correlation offset. */
#define CORR_MAX_BUF 63
#define CORR_DEV 4
#define CORR_MAX_LEVEL 20
#define CORR_MAX_LOW 4
#define CORR_BUF_LEN (CORR_MAX << 1) + 1
/* Note that CORR_WIDTH + 2*CORR_MAX <= MAX_BUF_LEN. */
#define ONE_Q14 (1 << 14)
/* NLP defines */
#define NLP_COMP_LOW 3277 /* 0.2 in Q14 */
#define NLP_COMP_HIGH ONE_Q14 /* 1 in Q14 */
#endif
sdk/lib/webrtc/modules/audio_processing/aecm/echo_control_mobile.c
+594
View File
@@ -0,0 +1,594 @@
/*
* Copyright (c) 2012 The WebRTC project authors. All Rights Reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
#include "echo_control_mobile.h"
#ifdef AEC_DEBUG
#include <stdio.h>
#endif
#include <stdlib.h>
#include <string.h>
#include "ring_buffer.h"
#include "signal_processing_library.h"
#include "aecm_defines.h"
#include "aecm_core.h"
#define BUF_SIZE_FRAMES 50 // buffer size (frames)
// Maximum length of resampled signal. Must be an integer multiple of frames
// (ceil(1/(1 + MIN_SKEW)*2) + 1)*FRAME_LEN
// The factor of 2 handles wb, and the + 1 is as a safety margin
#define MAX_RESAMP_LEN (5 * FRAME_LEN)
static const size_t kBufSizeSamp =
BUF_SIZE_FRAMES * FRAME_LEN; // buffer size (samples)
static const int kSampMsNb = 8; // samples per ms in nb
// Target suppression levels for nlp modes
// log{0.001, 0.00001, 0.00000001}
static const int kInitCheck = 42;
typedef struct {
int sampFreq;
int scSampFreq;
short bufSizeStart;
int knownDelay;
// Stores the last frame added to the farend buffer
short farendOld[2][FRAME_LEN];
short initFlag; // indicates if AEC has been initialized
// Variables used for averaging far end buffer size
short counter;
short sum;
short firstVal;
short checkBufSizeCtr;
// Variables used for delay shifts
short msInSndCardBuf;
short filtDelay;
int timeForDelayChange;
int ECstartup;
int checkBuffSize;
int delayChange;
short lastDelayDiff;
int16_t echoMode;
#ifdef AEC_DEBUG
FILE* bufFile;
FILE* delayFile;
FILE* preCompFile;
FILE* postCompFile;
#endif // AEC_DEBUG
// Structures
RingBuffer *farendBuf;
AecmCore *aecmCore;
} AecMobile;
// Estimates delay to set the position of the farend buffer read pointer
// (controlled by knownDelay)
static int WebRtcAecm_EstBufDelay(AecMobile *aecm, short msInSndCardBuf);
// Stuffs the farend buffer if the estimated delay is too large
static int WebRtcAecm_DelayComp(AecMobile *aecm);
void *WebRtcAecm_Create() {
// Allocate zero-filled memory.
AecMobile *aecm = (AecMobile *)(webrtc_calloc(1, sizeof(AecMobile)));
aecm->aecmCore = WebRtcAecm_CreateCore();
if (!aecm->aecmCore) {
WebRtcAecm_Free(aecm);
return NULL;
}
aecm->farendBuf = WebRtc_CreateBuffer(kBufSizeSamp, sizeof(int16_t));
if (!aecm->farendBuf) {
WebRtcAecm_Free(aecm);
return NULL;
}
#ifdef AEC_DEBUG
aecm->aecmCore->farFile = fopen("aecFar.pcm", "wb");
aecm->aecmCore->nearFile = fopen("aecNear.pcm", "wb");
aecm->aecmCore->outFile = fopen("aecOut.pcm", "wb");
// aecm->aecmCore->outLpFile = fopen("aecOutLp.pcm","wb");
aecm->bufFile = fopen("aecBuf.dat", "wb");
aecm->delayFile = fopen("aecDelay.dat", "wb");
aecm->preCompFile = fopen("preComp.pcm", "wb");
aecm->postCompFile = fopen("postComp.pcm", "wb");
#endif // AEC_DEBUG
return aecm;
}
void WebRtcAecm_Free(void *aecmInst) {
AecMobile *aecm = (AecMobile *)(aecmInst);
if (aecm == NULL) {
return;
}
#ifdef AEC_DEBUG
fclose(aecm->aecmCore->farFile);
fclose(aecm->aecmCore->nearFile);
fclose(aecm->aecmCore->outFile);
// fclose(aecm->aecmCore->outLpFile);
fclose(aecm->bufFile);
fclose(aecm->delayFile);
fclose(aecm->preCompFile);
fclose(aecm->postCompFile);
#endif // AEC_DEBUG
WebRtcAecm_FreeCore(aecm->aecmCore);
WebRtc_FreeBuffer(aecm->farendBuf);
webrtc_free(aecm);
}
int32_t WebRtcAecm_Init(void *aecmInst, int32_t sampFreq) {
AecMobile *aecm = (AecMobile *)(aecmInst);
AecmConfig aecConfig;
if (aecm == NULL) {
return -1;
}
if (sampFreq != 8000 && sampFreq != 16000) {
return AECM_BAD_PARAMETER_ERROR;
}
aecm->sampFreq = sampFreq;
// Initialize AECM core
if (WebRtcAecm_InitCore(aecm->aecmCore, aecm->sampFreq) == -1) {
return AECM_UNSPECIFIED_ERROR;
}
// Initialize farend buffer
WebRtc_InitBuffer(aecm->farendBuf);
aecm->initFlag = kInitCheck; // indicates that initialization has been done
aecm->delayChange = 1;
aecm->sum = 0;
aecm->counter = 0;
aecm->checkBuffSize = 1;
aecm->firstVal = 0;
aecm->ECstartup = 1;
aecm->bufSizeStart = 0;
aecm->checkBufSizeCtr = 0;
aecm->filtDelay = 0;
aecm->timeForDelayChange = 0;
aecm->knownDelay = 0;
aecm->lastDelayDiff = 0;
memset(&aecm->farendOld, 0, sizeof(aecm->farendOld));
// Default settings.
aecConfig.cngMode = AecmTrue;
aecConfig.echoMode = 3;
if (WebRtcAecm_set_config(aecm, aecConfig) == -1) {
return AECM_UNSPECIFIED_ERROR;
}
return 0;
}
// Returns any error that is caused when buffering the
// farend signal.
int32_t WebRtcAecm_GetBufferFarendError(void *aecmInst,
const int16_t *farend,
size_t nrOfSamples) {
AecMobile *aecm = (AecMobile *)(aecmInst);
if (aecm == NULL)
return -1;
if (farend == NULL)
return AECM_NULL_POINTER_ERROR;
if (aecm->initFlag != kInitCheck)
return AECM_UNINITIALIZED_ERROR;
if (nrOfSamples != 80 && nrOfSamples != 160)
return AECM_BAD_PARAMETER_ERROR;
return 0;
}
int32_t WebRtcAecm_BufferFarend(void *aecmInst,
const int16_t *farend,
size_t nrOfSamples) {
AecMobile *aecm = (AecMobile *)(aecmInst);
const int32_t err =
WebRtcAecm_GetBufferFarendError(aecmInst, farend, nrOfSamples);
if (err != 0)
return err;
// TODO(unknown): Is this really a good idea?
if (!aecm->ECstartup) {
WebRtcAecm_DelayComp(aecm);
}
WebRtc_WriteBuffer(aecm->farendBuf, farend, nrOfSamples);
return 0;
}
int32_t WebRtcAecm_Process(void *aecmInst,
const int16_t *nearendNoisy,
const int16_t *nearendClean,
int16_t *out,
size_t nrOfSamples,
int16_t msInSndCardBuf) {
AecMobile *aecm = (AecMobile *)(aecmInst);
int32_t retVal = 0;
size_t i;
volatile short nmbrOfFilledBuffers;
size_t nBlocks10ms;
size_t nFrames;
#ifdef AEC_DEBUG
short msInAECBuf;
#endif
if (aecm == NULL) {
return -1;
}
if (nearendNoisy == NULL) {
return AECM_NULL_POINTER_ERROR;
}
if (out == NULL) {
return AECM_NULL_POINTER_ERROR;
}
if (aecm->initFlag != kInitCheck) {
return AECM_UNINITIALIZED_ERROR;
}
if (nrOfSamples != 80 && nrOfSamples != 160) {
return AECM_BAD_PARAMETER_ERROR;
}
if (msInSndCardBuf < 0) {
msInSndCardBuf = 0;
retVal = AECM_BAD_PARAMETER_WARNING;
} else if (msInSndCardBuf > 500) {
msInSndCardBuf = 500;
retVal = AECM_BAD_PARAMETER_WARNING;
}
msInSndCardBuf += 10;
aecm->msInSndCardBuf = msInSndCardBuf;
nFrames = nrOfSamples / FRAME_LEN;
nBlocks10ms = nFrames / aecm->aecmCore->mult;
if (aecm->ECstartup) {
if (nearendClean == NULL) {
if (out != nearendNoisy) {
memcpy(out, nearendNoisy, sizeof(short) * nrOfSamples);
}
} else if (out != nearendClean) {
memcpy(out, nearendClean, sizeof(short) * nrOfSamples);
}
nmbrOfFilledBuffers =
(short) WebRtc_available_read(aecm->farendBuf) / FRAME_LEN;
// The AECM is in the start up mode
// AECM is disabled until the soundcard buffer and farend buffers are OK
// Mechanism to ensure that the soundcard buffer is reasonably stable.
if (aecm->checkBuffSize) {
aecm->checkBufSizeCtr++;
// Before we fill up the far end buffer we require the amount of data on
// the sound card to be stable (+/-8 ms) compared to the first value. This
// comparison is made during the following 4 consecutive frames. If it
// seems to be stable then we start to fill up the far end buffer.
if (aecm->counter == 0) {
aecm->firstVal = aecm->msInSndCardBuf;
aecm->sum = 0;
}
if (abs(aecm->firstVal - aecm->msInSndCardBuf) <
WEBRTC_SPL_MAX(0.2 * aecm->msInSndCardBuf, kSampMsNb)) {
aecm->sum += aecm->msInSndCardBuf;
aecm->counter++;
} else {
aecm->counter = 0;
}
if (aecm->counter * nBlocks10ms >= 6) {
// The farend buffer size is determined in blocks of 80 samples
// Use 75% of the average value of the soundcard buffer
aecm->bufSizeStart = WEBRTC_SPL_MIN(
(3 * aecm->sum * aecm->aecmCore->mult) / (aecm->counter * 40),
BUF_SIZE_FRAMES);
// buffersize has now been determined
aecm->checkBuffSize = 0;
}
if (aecm->checkBufSizeCtr * nBlocks10ms > 50) {
// for really bad sound cards, don't disable echocanceller for more than
// 0.5 sec
aecm->bufSizeStart = WEBRTC_SPL_MIN(
(3 * aecm->msInSndCardBuf * aecm->aecmCore->mult) / 40,
BUF_SIZE_FRAMES);
aecm->checkBuffSize = 0;
}
}
// if checkBuffSize changed in the if-statement above
if (!aecm->checkBuffSize) {
// soundcard buffer is now reasonably stable
// When the far end buffer is filled with approximately the same amount of
// data as the amount on the sound card we end the start up phase and
// start to cancel echoes.
if (nmbrOfFilledBuffers == aecm->bufSizeStart) {
aecm->ECstartup = 0; // Enable the AECM
} else if (nmbrOfFilledBuffers > aecm->bufSizeStart) {
WebRtc_MoveReadPtr(aecm->farendBuf,
(int) WebRtc_available_read(aecm->farendBuf) -
(int) aecm->bufSizeStart * FRAME_LEN);
aecm->ECstartup = 0;
}
}
} else {
// AECM is enabled
// Note only 1 block supported for nb and 2 blocks for wb
for (i = 0; i < nFrames; i++) {
int16_t farend[FRAME_LEN];
const int16_t *farend_ptr = NULL;
nmbrOfFilledBuffers =
(short) WebRtc_available_read(aecm->farendBuf) / FRAME_LEN;
// Check that there is data in the far end buffer
if (nmbrOfFilledBuffers > 0) {
// Get the next 80 samples from the farend buffer
WebRtc_ReadBuffer(aecm->farendBuf, (void **) &farend_ptr, farend,
FRAME_LEN);
// Always store the last frame for use when we run out of data
memcpy(&(aecm->farendOld[i][0]), farend_ptr, FRAME_LEN * sizeof(short));
} else {
// We have no data so we use the last played frame
memcpy(farend, &(aecm->farendOld[i][0]), FRAME_LEN * sizeof(short));
farend_ptr = farend;
}
// Call buffer delay estimator when all data is extracted,
// i,e. i = 0 for NB and i = 1 for WB
if ((i == 0 && aecm->sampFreq == 8000) ||
(i == 1 && aecm->sampFreq == 16000)) {
WebRtcAecm_EstBufDelay(aecm, aecm->msInSndCardBuf);
}
// Call the AECM
/*WebRtcAecm_ProcessFrame(aecm->aecmCore, farend, &nearend[FRAME_LEN * i],
&out[FRAME_LEN * i], aecm->knownDelay);*/
if (WebRtcAecm_ProcessFrame(
aecm->aecmCore, farend_ptr, &nearendNoisy[FRAME_LEN * i],
(nearendClean ? &nearendClean[FRAME_LEN * i] : NULL),
&out[FRAME_LEN * i]) == -1)
return -1;
}
}
#ifdef AEC_DEBUG
msInAECBuf = (short)WebRtc_available_read(aecm->farendBuf) /
(kSampMsNb * aecm->aecmCore->mult);
fwrite(&msInAECBuf, 2, 1, aecm->bufFile);
fwrite(&(aecm->knownDelay), sizeof(aecm->knownDelay), 1, aecm->delayFile);
#endif
return retVal;
}
int32_t WebRtcAecm_set_config(void *aecmInst, AecmConfig config) {
AecMobile *aecm = (AecMobile *)(aecmInst);
if (aecm == NULL) {
return -1;
}
if (aecm->initFlag != kInitCheck) {
return AECM_UNINITIALIZED_ERROR;
}
if (config.cngMode != AecmFalse && config.cngMode != AecmTrue) {
return AECM_BAD_PARAMETER_ERROR;
}
aecm->aecmCore->cngMode = config.cngMode;
if (config.echoMode < 0 || config.echoMode > 4) {
return AECM_BAD_PARAMETER_ERROR;
}
aecm->echoMode = config.echoMode;
if (aecm->echoMode == 0) {
aecm->aecmCore->supGain = SUPGAIN_DEFAULT >> 3;
aecm->aecmCore->supGainOld = SUPGAIN_DEFAULT >> 3;
aecm->aecmCore->supGainErrParamA = SUPGAIN_ERROR_PARAM_A >> 3;
aecm->aecmCore->supGainErrParamD = SUPGAIN_ERROR_PARAM_D >> 3;
aecm->aecmCore->supGainErrParamDiffAB =
(SUPGAIN_ERROR_PARAM_A >> 3) - (SUPGAIN_ERROR_PARAM_B >> 3);
aecm->aecmCore->supGainErrParamDiffBD =
(SUPGAIN_ERROR_PARAM_B >> 3) - (SUPGAIN_ERROR_PARAM_D >> 3);
} else if (aecm->echoMode == 1) {
aecm->aecmCore->supGain = SUPGAIN_DEFAULT >> 2;
aecm->aecmCore->supGainOld = SUPGAIN_DEFAULT >> 2;
aecm->aecmCore->supGainErrParamA = SUPGAIN_ERROR_PARAM_A >> 2;
aecm->aecmCore->supGainErrParamD = SUPGAIN_ERROR_PARAM_D >> 2;
aecm->aecmCore->supGainErrParamDiffAB =
(SUPGAIN_ERROR_PARAM_A >> 2) - (SUPGAIN_ERROR_PARAM_B >> 2);
aecm->aecmCore->supGainErrParamDiffBD =
(SUPGAIN_ERROR_PARAM_B >> 2) - (SUPGAIN_ERROR_PARAM_D >> 2);
} else if (aecm->echoMode == 2) {
aecm->aecmCore->supGain = SUPGAIN_DEFAULT >> 1;
aecm->aecmCore->supGainOld = SUPGAIN_DEFAULT >> 1;
aecm->aecmCore->supGainErrParamA = SUPGAIN_ERROR_PARAM_A >> 1;
aecm->aecmCore->supGainErrParamD = SUPGAIN_ERROR_PARAM_D >> 1;
aecm->aecmCore->supGainErrParamDiffAB =
(SUPGAIN_ERROR_PARAM_A >> 1) - (SUPGAIN_ERROR_PARAM_B >> 1);
aecm->aecmCore->supGainErrParamDiffBD =
(SUPGAIN_ERROR_PARAM_B >> 1) - (SUPGAIN_ERROR_PARAM_D >> 1);
} else if (aecm->echoMode == 3) {
aecm->aecmCore->supGain = SUPGAIN_DEFAULT;
aecm->aecmCore->supGainOld = SUPGAIN_DEFAULT;
aecm->aecmCore->supGainErrParamA = SUPGAIN_ERROR_PARAM_A;
aecm->aecmCore->supGainErrParamD = SUPGAIN_ERROR_PARAM_D;
aecm->aecmCore->supGainErrParamDiffAB =
SUPGAIN_ERROR_PARAM_A - SUPGAIN_ERROR_PARAM_B;
aecm->aecmCore->supGainErrParamDiffBD =
SUPGAIN_ERROR_PARAM_B - SUPGAIN_ERROR_PARAM_D;
} else if (aecm->echoMode == 4) {
aecm->aecmCore->supGain = SUPGAIN_DEFAULT << 1;
aecm->aecmCore->supGainOld = SUPGAIN_DEFAULT << 1;
aecm->aecmCore->supGainErrParamA = SUPGAIN_ERROR_PARAM_A << 1;
aecm->aecmCore->supGainErrParamD = SUPGAIN_ERROR_PARAM_D << 1;
aecm->aecmCore->supGainErrParamDiffAB =
(SUPGAIN_ERROR_PARAM_A << 1) - (SUPGAIN_ERROR_PARAM_B << 1);
aecm->aecmCore->supGainErrParamDiffBD =
(SUPGAIN_ERROR_PARAM_B << 1) - (SUPGAIN_ERROR_PARAM_D << 1);
}
return 0;
}
int32_t WebRtcAecm_InitEchoPath(void *aecmInst,
const void *echo_path,
size_t size_bytes) {
AecMobile *aecm = (AecMobile *)(aecmInst);
const int16_t *echo_path_ptr = (const int16_t *)(echo_path);
if (aecmInst == NULL) {
return -1;
}
if (echo_path == NULL) {
return AECM_NULL_POINTER_ERROR;
}
if (size_bytes != WebRtcAecm_echo_path_size_bytes()) {
// Input channel size does not match the size of AECM
return AECM_BAD_PARAMETER_ERROR;
}
if (aecm->initFlag != kInitCheck) {
return AECM_UNINITIALIZED_ERROR;
}
WebRtcAecm_InitEchoPathCore(aecm->aecmCore, echo_path_ptr);
return 0;
}
int32_t WebRtcAecm_GetEchoPath(void *aecmInst,
void *echo_path,
size_t size_bytes) {
AecMobile *aecm = (AecMobile *)(aecmInst);
int16_t *echo_path_ptr = (int16_t *)(echo_path);
if (aecmInst == NULL) {
return -1;
}
if (echo_path == NULL) {
return AECM_NULL_POINTER_ERROR;
}
if (size_bytes != WebRtcAecm_echo_path_size_bytes()) {
// Input channel size does not match the size of AECM
return AECM_BAD_PARAMETER_ERROR;
}
if (aecm->initFlag != kInitCheck) {
return AECM_UNINITIALIZED_ERROR;
}
memcpy(echo_path_ptr, aecm->aecmCore->channelStored, size_bytes);
return 0;
}
size_t WebRtcAecm_echo_path_size_bytes() {
return (PART_LEN1 * sizeof(int16_t));
}
static int WebRtcAecm_EstBufDelay(AecMobile *aecm, short msInSndCardBuf) {
short delayNew, nSampSndCard;
short nSampFar = (short) WebRtc_available_read(aecm->farendBuf);
short diff;
nSampSndCard = msInSndCardBuf * kSampMsNb * aecm->aecmCore->mult;
delayNew = nSampSndCard - nSampFar;
if (delayNew < FRAME_LEN) {
WebRtc_MoveReadPtr(aecm->farendBuf, FRAME_LEN);
delayNew += FRAME_LEN;
}
aecm->filtDelay =
WEBRTC_SPL_MAX(0, (8 * aecm->filtDelay + 2 * delayNew) / 10);
diff = aecm->filtDelay - aecm->knownDelay;
if (diff > 224) {
if (aecm->lastDelayDiff < 96) {
aecm->timeForDelayChange = 0;
} else {
aecm->timeForDelayChange++;
}
} else if (diff < 96 && aecm->knownDelay > 0) {
if (aecm->lastDelayDiff > 224) {
aecm->timeForDelayChange = 0;
} else {
aecm->timeForDelayChange++;
}
} else {
aecm->timeForDelayChange = 0;
}
aecm->lastDelayDiff = diff;
if (aecm->timeForDelayChange > 25) {
aecm->knownDelay = WEBRTC_SPL_MAX((int) aecm->filtDelay - 160, 0);
}
return 0;
}
static int WebRtcAecm_DelayComp(AecMobile *aecm) {
int nSampFar = (int) WebRtc_available_read(aecm->farendBuf);
int nSampSndCard, delayNew, nSampAdd;
const int maxStuffSamp = 10 * FRAME_LEN;
nSampSndCard = aecm->msInSndCardBuf * kSampMsNb * aecm->aecmCore->mult;
delayNew = nSampSndCard - nSampFar;
if (delayNew > FAR_BUF_LEN - FRAME_LEN * aecm->aecmCore->mult) {
// The difference of the buffer sizes is larger than the maximum
// allowed known delay. Compensate by stuffing the buffer.
nSampAdd = (WEBRTC_SPL_MAX(((nSampSndCard >> 1) - nSampFar), FRAME_LEN));
nSampAdd = WEBRTC_SPL_MIN(nSampAdd, maxStuffSamp);
WebRtc_MoveReadPtr(aecm->farendBuf, -nSampAdd);
aecm->delayChange = 1; // the delay needs to be updated
}
return 0;
}
sdk/lib/webrtc/modules/audio_processing/aecm/echo_control_mobile.h
+201
View File
@@ -0,0 +1,201 @@
/*
* Copyright (c) 2012 The WebRTC project authors. All Rights Reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
#ifndef MODULES_AUDIO_PROCESSING_AECM_ECHO_CONTROL_MOBILE_H_
#define MODULES_AUDIO_PROCESSING_AECM_ECHO_CONTROL_MOBILE_H_
#include <stddef.h>
#include "webrtc_typedef.h"
enum {
AecmFalse = 0, AecmTrue
};
// Errors
#define AECM_UNSPECIFIED_ERROR 12000
#define AECM_UNSUPPORTED_FUNCTION_ERROR 12001
#define AECM_UNINITIALIZED_ERROR 12002
#define AECM_NULL_POINTER_ERROR 12003
#define AECM_BAD_PARAMETER_ERROR 12004
// Warnings
#define AECM_BAD_PARAMETER_WARNING 12100
typedef struct {
int16_t cngMode; // AECM_FALSE, AECM_TRUE (default)
int16_t echoMode; // 0, 1, 2, 3 (default), 4
} AecmConfig;
/*
* Allocates the memory needed by the AECM. The memory needs to be
* initialized separately using the WebRtcAecm_Init() function.
* Returns a pointer to the instance and a nullptr at failure.
*/
void *WebRtcAecm_Create();
/*
* This function releases the memory allocated by WebRtcAecm_Create()
*
* Inputs Description
* -------------------------------------------------------------------
* void* aecmInst Pointer to the AECM instance
*/
void WebRtcAecm_Free(void *aecmInst);
/*
* Initializes an AECM instance.
*
* Inputs Description
* -------------------------------------------------------------------
* void* aecmInst Pointer to the AECM instance
* int32_t sampFreq Sampling frequency of data
*
* Outputs Description
* -------------------------------------------------------------------
* int32_t return 0: OK
* 1200-12004,12100: error/warning
*/
int32_t WebRtcAecm_Init(void *aecmInst, int32_t sampFreq);
/*
* Inserts an 80 or 160 sample block of data into the farend buffer.
*
* Inputs Description
* -------------------------------------------------------------------
* void* aecmInst Pointer to the AECM instance
* int16_t* farend In buffer containing one frame of
* farend signal
* int16_t nrOfSamples Number of samples in farend buffer
*
* Outputs Description
* -------------------------------------------------------------------
* int32_t return 0: OK
* 1200-12004,12100: error/warning
*/
int32_t WebRtcAecm_BufferFarend(void *aecmInst,
const int16_t *farend,
size_t nrOfSamples);
/*
* Reports any errors that would arise when buffering a farend buffer.
*
* Inputs Description
* -------------------------------------------------------------------
* void* aecmInst Pointer to the AECM instance
* int16_t* farend In buffer containing one frame of
* farend signal
* int16_t nrOfSamples Number of samples in farend buffer
*
* Outputs Description
* -------------------------------------------------------------------
* int32_t return 0: OK
* 1200-12004,12100: error/warning
*/
int32_t WebRtcAecm_GetBufferFarendError(void *aecmInst,
const int16_t *farend,
size_t nrOfSamples);
/*
* Runs the AECM on an 80 or 160 sample blocks of data.
*
* Inputs Description
* -------------------------------------------------------------------
* void* aecmInst Pointer to the AECM instance
* int16_t* nearendNoisy In buffer containing one frame of
* reference nearend+echo signal. If
* noise reduction is active, provide
* the noisy signal here.
* int16_t* nearendClean In buffer containing one frame of
* nearend+echo signal. If noise
* reduction is active, provide the
* clean signal here. Otherwise pass a
* NULL pointer.
* int16_t nrOfSamples Number of samples in nearend buffer
* int16_t msInSndCardBuf Delay estimate for sound card and
* system buffers
*
* Outputs Description
* -------------------------------------------------------------------
* int16_t* out Out buffer, one frame of processed nearend
* int32_t return 0: OK
* 1200-12004,12100: error/warning
*/
int32_t WebRtcAecm_Process(void *aecmInst,
const int16_t *nearendNoisy,
const int16_t *nearendClean,
int16_t *out,
size_t nrOfSamples,
int16_t msInSndCardBuf);
/*
* This function enables the user to set certain parameters on-the-fly
*
* Inputs Description
* -------------------------------------------------------------------
* void* aecmInst Pointer to the AECM instance
* AecmConfig config Config instance that contains all
* properties to be set
*
* Outputs Description
* -------------------------------------------------------------------
* int32_t return 0: OK
* 1200-12004,12100: error/warning
*/
int32_t WebRtcAecm_set_config(void *aecmInst, AecmConfig config);
/*
* This function enables the user to set the echo path on-the-fly.
*
* Inputs Description
* -------------------------------------------------------------------
* void* aecmInst Pointer to the AECM instance
* void* echo_path Pointer to the echo path to be set
* size_t size_bytes Size in bytes of the echo path
*
* Outputs Description
* -------------------------------------------------------------------
* int32_t return 0: OK
* 1200-12004,12100: error/warning
*/
int32_t WebRtcAecm_InitEchoPath(void *aecmInst,
const void *echo_path,
size_t size_bytes);
/*
* This function enables the user to get the currently used echo path
* on-the-fly
*
* Inputs Description
* -------------------------------------------------------------------
* void* aecmInst Pointer to the AECM instance
* void* echo_path Pointer to echo path
* size_t size_bytes Size in bytes of the echo path
*
* Outputs Description
* -------------------------------------------------------------------
* int32_t return 0: OK
* 1200-12004,12100: error/warning
*/
int32_t WebRtcAecm_GetEchoPath(void *aecmInst,
void *echo_path,
size_t size_bytes);
/*
* This function enables the user to get the echo path size in bytes
*
* Outputs Description
* -------------------------------------------------------------------
* size_t return Size in bytes
*/
size_t WebRtcAecm_echo_path_size_bytes();
#endif // MODULES_AUDIO_PROCESSING_AECM_ECHO_CONTROL_MOBILE_H_