Avoiding overflow in cross correlation in NetEq.
BUG=
Review-Url: https://codereview.webrtc.org/1908623002
Cr-Commit-Position: refs/heads/master@{#12538}
diff --git a/webrtc/modules/audio_coding/neteq/expand.cc b/webrtc/modules/audio_coding/neteq/expand.cc
index ef7af46..94f6a8f 100644
--- a/webrtc/modules/audio_coding/neteq/expand.cc
+++ b/webrtc/modules/audio_coding/neteq/expand.cc
@@ -19,6 +19,7 @@
#include "webrtc/base/safe_conversions.h"
#include "webrtc/common_audio/signal_processing/include/signal_processing_library.h"
#include "webrtc/modules/audio_coding/neteq/background_noise.h"
+#include "webrtc/modules/audio_coding/neteq/cross_correlation.h"
#include "webrtc/modules/audio_coding/neteq/dsp_helper.h"
#include "webrtc/modules/audio_coding/neteq/random_vector.h"
#include "webrtc/modules/audio_coding/neteq/statistics_calculator.h"
@@ -379,12 +380,10 @@
InitializeForAnExpandPeriod();
// Calculate correlation in downsampled domain (4 kHz sample rate).
- int correlation_scale;
size_t correlation_length = 51; // TODO(hlundin): Legacy bit-exactness.
// If it is decided to break bit-exactness |correlation_length| should be
// initialized to the return value of Correlation().
- Correlation(audio_history, signal_length, correlation_vector,
- &correlation_scale);
+ Correlation(audio_history, signal_length, correlation_vector);
// Find peaks in correlation vector.
DspHelper::PeakDetection(correlation_vector, correlation_length,
@@ -450,21 +449,12 @@
for (size_t channel_ix = 0; channel_ix < num_channels_; ++channel_ix) {
ChannelParameters& parameters = channel_parameters_[channel_ix];
- // Calculate suitable scaling.
- int16_t signal_max = WebRtcSpl_MaxAbsValueW16(
- &audio_history[signal_length - correlation_length - start_index
- - correlation_lags],
- correlation_length + start_index + correlation_lags - 1);
- correlation_scale = (31 - WebRtcSpl_NormW32(signal_max * signal_max)) +
- (31 - WebRtcSpl_NormW32(static_cast<int32_t>(correlation_length))) - 31;
- correlation_scale = std::max(0, correlation_scale);
// Calculate the correlation, store in |correlation_vector2|.
- WebRtcSpl_CrossCorrelation(
- correlation_vector2,
+ int correlation_scale = CrossCorrelationWithAutoShift(
&(audio_history[signal_length - correlation_length]),
&(audio_history[signal_length - correlation_length - start_index]),
- correlation_length, correlation_lags, correlation_scale, -1);
+ correlation_length, correlation_lags, -1, correlation_vector2);
// Find maximizing index.
best_index = WebRtcSpl_MaxIndexW32(correlation_vector2, correlation_lags);
@@ -582,13 +572,6 @@
}
// Calculate the LPC and the gain of the filters.
- // Calculate scale value needed for auto-correlation.
- correlation_scale = WebRtcSpl_MaxAbsValueW16(
- &(audio_history[signal_length - fs_mult_lpc_analysis_len]),
- fs_mult_lpc_analysis_len);
-
- correlation_scale = std::min(16 - WebRtcSpl_NormW32(correlation_scale), 0);
- correlation_scale = std::max(correlation_scale * 2 + 7, 0);
// Calculate kUnvoicedLpcOrder + 1 lags of the auto-correlation function.
size_t temp_index = signal_length - fs_mult_lpc_analysis_len -
@@ -601,11 +584,9 @@
memcpy(&temp_signal[kUnvoicedLpcOrder],
&audio_history[temp_index + kUnvoicedLpcOrder],
sizeof(int16_t) * fs_mult_lpc_analysis_len);
- WebRtcSpl_CrossCorrelation(auto_correlation,
- &temp_signal[kUnvoicedLpcOrder],
- &temp_signal[kUnvoicedLpcOrder],
- fs_mult_lpc_analysis_len, kUnvoicedLpcOrder + 1,
- correlation_scale, -1);
+ correlation_scale = CrossCorrelationWithAutoShift(
+ &temp_signal[kUnvoicedLpcOrder], &temp_signal[kUnvoicedLpcOrder],
+ fs_mult_lpc_analysis_len, kUnvoicedLpcOrder + 1, -1, auto_correlation);
delete [] temp_signal;
// Verify that variance is positive.
@@ -766,8 +747,7 @@
void Expand::Correlation(const int16_t* input,
size_t input_length,
- int16_t* output,
- int* output_scale) const {
+ int16_t* output) const {
// Set parameters depending on sample rate.
const int16_t* filter_coefficients;
size_t num_coefficients;
@@ -814,13 +794,11 @@
downsampled_input, norm_shift);
int32_t correlation[kNumCorrelationLags];
- static const int kCorrelationShift = 6;
- WebRtcSpl_CrossCorrelation(
- correlation,
+ CrossCorrelationWithAutoShift(
&downsampled_input[kDownsampledLength - kCorrelationLength],
&downsampled_input[kDownsampledLength - kCorrelationLength
- kCorrelationStartLag],
- kCorrelationLength, kNumCorrelationLags, kCorrelationShift, -1);
+ kCorrelationLength, kNumCorrelationLags, -1, correlation);
// Normalize and move data from 32-bit to 16-bit vector.
int32_t max_correlation = WebRtcSpl_MaxAbsValueW32(correlation,
@@ -829,8 +807,6 @@
std::max(18 - WebRtcSpl_NormW32(max_correlation), 0));
WebRtcSpl_VectorBitShiftW32ToW16(output, kNumCorrelationLags, correlation,
norm_shift2);
- // Total scale factor (right shifts) of correlation value.
- *output_scale = 2 * norm_shift + kCorrelationShift + norm_shift2;
}
void Expand::UpdateLagIndex() {