Home | History | Annotate | Download | only in voice_engine 
      1 /*
      2  *  Copyright (c) 2012 The WebRTC project authors. All Rights Reserved.
      3  *
      4  *  Use of this source code is governed by a BSD-style license
      5  *  that can be found in the LICENSE file in the root of the source
      6  *  tree. An additional intellectual property rights grant can be found
      7  *  in the file PATENTS.  All contributing project authors may
      8  *  be found in the AUTHORS file in the root of the source tree.
      9  */
     10 
     11 #include <math.h>
     12 
     13 #include "testing/gtest/include/gtest/gtest.h"
     14 #include "webrtc/base/format_macros.h"
     15 #include "webrtc/common_audio/resampler/include/push_resampler.h"
     16 #include "webrtc/modules/include/module_common_types.h"
     17 #include "webrtc/voice_engine/utility.h"
     18 #include "webrtc/voice_engine/voice_engine_defines.h"
     19 
     20 namespace webrtc {
     21 namespace voe {
     22 namespace {
     23 
     24 class UtilityTest : public ::testing::Test {
     25  protected:
     26   UtilityTest() {
     27     src_frame_.sample_rate_hz_ = 16000;
     28     src_frame_.samples_per_channel_ = src_frame_.sample_rate_hz_ / 100;
     29     src_frame_.num_channels_ = 1;
     30     dst_frame_.CopyFrom(src_frame_);
     31     golden_frame_.CopyFrom(src_frame_);
     32   }
     33 
     34   void RunResampleTest(int src_channels,
     35                        int src_sample_rate_hz,
     36                        int dst_channels,
     37                        int dst_sample_rate_hz);
     38 
     39   PushResampler<int16_t> resampler_;
     40   AudioFrame src_frame_;
     41   AudioFrame dst_frame_;
     42   AudioFrame golden_frame_;
     43 };
     44 
     45 // Sets the signal value to increase by |data| with every sample. Floats are
     46 // used so non-integer values result in rounding error, but not an accumulating
     47 // error.
     48 void SetMonoFrame(AudioFrame* frame, float data, int sample_rate_hz) {
     49   memset(frame->data_, 0, sizeof(frame->data_));
     50   frame->num_channels_ = 1;
     51   frame->sample_rate_hz_ = sample_rate_hz;
     52   frame->samples_per_channel_ = sample_rate_hz / 100;
     53   for (size_t i = 0; i < frame->samples_per_channel_; i++) {
     54     frame->data_[i] = static_cast<int16_t>(data * i);
     55   }
     56 }
     57 
     58 // Keep the existing sample rate.
     59 void SetMonoFrame(AudioFrame* frame, float data) {
     60   SetMonoFrame(frame, data, frame->sample_rate_hz_);
     61 }
     62 
     63 // Sets the signal value to increase by |left| and |right| with every sample in
     64 // each channel respectively.
     65 void SetStereoFrame(AudioFrame* frame, float left, float right,
     66                     int sample_rate_hz) {
     67   memset(frame->data_, 0, sizeof(frame->data_));
     68   frame->num_channels_ = 2;
     69   frame->sample_rate_hz_ = sample_rate_hz;
     70   frame->samples_per_channel_ = sample_rate_hz / 100;
     71   for (size_t i = 0; i < frame->samples_per_channel_; i++) {
     72     frame->data_[i * 2] = static_cast<int16_t>(left * i);
     73     frame->data_[i * 2 + 1] = static_cast<int16_t>(right * i);
     74   }
     75 }
     76 
     77 // Keep the existing sample rate.
     78 void SetStereoFrame(AudioFrame* frame, float left, float right) {
     79   SetStereoFrame(frame, left, right, frame->sample_rate_hz_);
     80 }
     81 
     82 void VerifyParams(const AudioFrame& ref_frame, const AudioFrame& test_frame) {
     83   EXPECT_EQ(ref_frame.num_channels_, test_frame.num_channels_);
     84   EXPECT_EQ(ref_frame.samples_per_channel_, test_frame.samples_per_channel_);
     85   EXPECT_EQ(ref_frame.sample_rate_hz_, test_frame.sample_rate_hz_);
     86 }
     87 
     88 // Computes the best SNR based on the error between |ref_frame| and
     89 // |test_frame|. It allows for up to a |max_delay| in samples between the
     90 // signals to compensate for the resampling delay.
     91 float ComputeSNR(const AudioFrame& ref_frame, const AudioFrame& test_frame,
     92                  size_t max_delay) {
     93   VerifyParams(ref_frame, test_frame);
     94   float best_snr = 0;
     95   size_t best_delay = 0;
     96   for (size_t delay = 0; delay <= max_delay; delay++) {
     97     float mse = 0;
     98     float variance = 0;
     99     for (size_t i = 0; i < ref_frame.samples_per_channel_ *
    100         ref_frame.num_channels_ - delay; i++) {
    101       int error = ref_frame.data_[i] - test_frame.data_[i + delay];
    102       mse += error * error;
    103       variance += ref_frame.data_[i] * ref_frame.data_[i];
    104     }
    105     float snr = 100;  // We assign 100 dB to the zero-error case.
    106     if (mse > 0)
    107       snr = 10 * log10(variance / mse);
    108     if (snr > best_snr) {
    109       best_snr = snr;
    110       best_delay = delay;
    111     }
    112   }
    113   printf("SNR=%.1f dB at delay=%" PRIuS "\n", best_snr, best_delay);
    114   return best_snr;
    115 }
    116 
    117 void VerifyFramesAreEqual(const AudioFrame& ref_frame,
    118                           const AudioFrame& test_frame) {
    119   VerifyParams(ref_frame, test_frame);
    120   for (size_t i = 0;
    121        i < ref_frame.samples_per_channel_ * ref_frame.num_channels_; i++) {
    122     EXPECT_EQ(ref_frame.data_[i], test_frame.data_[i]);
    123   }
    124 }
    125 
    126 void UtilityTest::RunResampleTest(int src_channels,
    127                                   int src_sample_rate_hz,
    128                                   int dst_channels,
    129                                   int dst_sample_rate_hz) {
    130   PushResampler<int16_t> resampler;  // Create a new one with every test.
    131   const int16_t kSrcLeft = 30;  // Shouldn't overflow for any used sample rate.
    132   const int16_t kSrcRight = 15;
    133   const float resampling_factor = (1.0 * src_sample_rate_hz) /
    134       dst_sample_rate_hz;
    135   const float dst_left = resampling_factor * kSrcLeft;
    136   const float dst_right = resampling_factor * kSrcRight;
    137   const float dst_mono = (dst_left + dst_right) / 2;
    138   if (src_channels == 1)
    139     SetMonoFrame(&src_frame_, kSrcLeft, src_sample_rate_hz);
    140   else
    141     SetStereoFrame(&src_frame_, kSrcLeft, kSrcRight, src_sample_rate_hz);
    142 
    143   if (dst_channels == 1) {
    144     SetMonoFrame(&dst_frame_, 0, dst_sample_rate_hz);
    145     if (src_channels == 1)
    146       SetMonoFrame(&golden_frame_, dst_left, dst_sample_rate_hz);
    147     else
    148       SetMonoFrame(&golden_frame_, dst_mono, dst_sample_rate_hz);
    149   } else {
    150     SetStereoFrame(&dst_frame_, 0, 0, dst_sample_rate_hz);
    151     if (src_channels == 1)
    152       SetStereoFrame(&golden_frame_, dst_left, dst_left, dst_sample_rate_hz);
    153     else
    154       SetStereoFrame(&golden_frame_, dst_left, dst_right, dst_sample_rate_hz);
    155   }
    156 
    157   // The sinc resampler has a known delay, which we compute here. Multiplying by
    158   // two gives us a crude maximum for any resampling, as the old resampler
    159   // typically (but not always) has lower delay.
    160   static const size_t kInputKernelDelaySamples = 16;
    161   const size_t max_delay = static_cast<size_t>(
    162       static_cast<double>(dst_sample_rate_hz) / src_sample_rate_hz *
    163       kInputKernelDelaySamples * dst_channels * 2);
    164   printf("(%d, %d Hz) -> (%d, %d Hz) ",  // SNR reported on the same line later.
    165       src_channels, src_sample_rate_hz, dst_channels, dst_sample_rate_hz);
    166   RemixAndResample(src_frame_, &resampler, &dst_frame_);
    167 
    168   if (src_sample_rate_hz == 96000 && dst_sample_rate_hz == 8000) {
    169     // The sinc resampler gives poor SNR at this extreme conversion, but we
    170     // expect to see this rarely in practice.
    171     EXPECT_GT(ComputeSNR(golden_frame_, dst_frame_, max_delay), 14.0f);
    172   } else {
    173     EXPECT_GT(ComputeSNR(golden_frame_, dst_frame_, max_delay), 46.0f);
    174   }
    175 }
    176 
    177 TEST_F(UtilityTest, RemixAndResampleCopyFrameSucceeds) {
    178   // Stereo -> stereo.
    179   SetStereoFrame(&src_frame_, 10, 10);
    180   SetStereoFrame(&dst_frame_, 0, 0);
    181   RemixAndResample(src_frame_, &resampler_, &dst_frame_);
    182   VerifyFramesAreEqual(src_frame_, dst_frame_);
    183 
    184   // Mono -> mono.
    185   SetMonoFrame(&src_frame_, 20);
    186   SetMonoFrame(&dst_frame_, 0);
    187   RemixAndResample(src_frame_, &resampler_, &dst_frame_);
    188   VerifyFramesAreEqual(src_frame_, dst_frame_);
    189 }
    190 
    191 TEST_F(UtilityTest, RemixAndResampleMixingOnlySucceeds) {
    192   // Stereo -> mono.
    193   SetStereoFrame(&dst_frame_, 0, 0);
    194   SetMonoFrame(&src_frame_, 10);
    195   SetStereoFrame(&golden_frame_, 10, 10);
    196   RemixAndResample(src_frame_, &resampler_, &dst_frame_);
    197   VerifyFramesAreEqual(dst_frame_, golden_frame_);
    198 
    199   // Mono -> stereo.
    200   SetMonoFrame(&dst_frame_, 0);
    201   SetStereoFrame(&src_frame_, 10, 20);
    202   SetMonoFrame(&golden_frame_, 15);
    203   RemixAndResample(src_frame_, &resampler_, &dst_frame_);
    204   VerifyFramesAreEqual(golden_frame_, dst_frame_);
    205 }
    206 
    207 TEST_F(UtilityTest, RemixAndResampleSucceeds) {
    208   const int kSampleRates[] = {8000, 16000, 32000, 44100, 48000, 96000};
    209   const int kSampleRatesSize = sizeof(kSampleRates) / sizeof(*kSampleRates);
    210   const int kChannels[] = {1, 2};
    211   const int kChannelsSize = sizeof(kChannels) / sizeof(*kChannels);
    212   for (int src_rate = 0; src_rate < kSampleRatesSize; src_rate++) {
    213     for (int dst_rate = 0; dst_rate < kSampleRatesSize; dst_rate++) {
    214       for (int src_channel = 0; src_channel < kChannelsSize; src_channel++) {
    215         for (int dst_channel = 0; dst_channel < kChannelsSize; dst_channel++) {
    216           RunResampleTest(kChannels[src_channel], kSampleRates[src_rate],
    217                           kChannels[dst_channel], kSampleRates[dst_rate]);
    218         }
    219       }
    220     }
    221   }
    222 }
    223 
    224 }  // namespace
    225 }  // namespace voe
    226 }  // namespace webrtc
    227