Home | History | Annotate | Download | only in vad 
      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 <stdlib.h>
     12 
     13 #include "testing/gtest/include/gtest/gtest.h"
     14 #include "webrtc/common_audio/vad/vad_unittest.h"
     15 #include "webrtc/typedefs.h"
     16 
     17 extern "C" {
     18 #include "webrtc/common_audio/vad/vad_core.h"
     19 #include "webrtc/common_audio/vad/vad_filterbank.h"
     20 }
     21 
     22 namespace {
     23 
     24 const int kNumValidFrameLengths = 3;
     25 
     26 TEST_F(VadTest, vad_filterbank) {
     27   VadInstT* self = reinterpret_cast<VadInstT*>(malloc(sizeof(VadInstT)));
     28   static const int16_t kReference[kNumValidFrameLengths] = { 48, 11, 11 };
     29   static const int16_t kFeatures[kNumValidFrameLengths * kNumChannels] = {
     30       1213, 759, 587, 462, 434, 272,
     31       1479, 1385, 1291, 1200, 1103, 1099,
     32       1732, 1692, 1681, 1629, 1436, 1436
     33   };
     34   static const int16_t kOffsetVector[kNumChannels] = {
     35       368, 368, 272, 176, 176, 176 };
     36   int16_t features[kNumChannels];
     37 
     38   // Construct a speech signal that will trigger the VAD in all modes. It is
     39   // known that (i * i) will wrap around, but that doesn't matter in this case.
     40   int16_t speech[kMaxFrameLength];
     41   for (int16_t i = 0; i < kMaxFrameLength; ++i) {
     42     speech[i] = (i * i);
     43   }
     44 
     45   int frame_length_index = 0;
     46   ASSERT_EQ(0, WebRtcVad_InitCore(self));
     47   for (size_t j = 0; j < kFrameLengthsSize; ++j) {
     48     if (ValidRatesAndFrameLengths(8000, kFrameLengths[j])) {
     49       EXPECT_EQ(kReference[frame_length_index],
     50                 WebRtcVad_CalculateFeatures(self, speech, kFrameLengths[j],
     51                                             features));
     52       for (int k = 0; k < kNumChannels; ++k) {
     53         EXPECT_EQ(kFeatures[k + frame_length_index * kNumChannels],
     54                   features[k]);
     55       }
     56       frame_length_index++;
     57     }
     58   }
     59   EXPECT_EQ(kNumValidFrameLengths, frame_length_index);
     60 
     61   // Verify that all zeros in gives kOffsetVector out.
     62   memset(speech, 0, sizeof(speech));
     63   ASSERT_EQ(0, WebRtcVad_InitCore(self));
     64   for (size_t j = 0; j < kFrameLengthsSize; ++j) {
     65     if (ValidRatesAndFrameLengths(8000, kFrameLengths[j])) {
     66       EXPECT_EQ(0, WebRtcVad_CalculateFeatures(self, speech, kFrameLengths[j],
     67                                                features));
     68       for (int k = 0; k < kNumChannels; ++k) {
     69         EXPECT_EQ(kOffsetVector[k], features[k]);
     70       }
     71     }
     72   }
     73 
     74   // Verify that all ones in gives kOffsetVector out. Any other constant input
     75   // will have a small impact in the sub bands.
     76   for (int16_t i = 0; i < kMaxFrameLength; ++i) {
     77     speech[i] = 1;
     78   }
     79   for (size_t j = 0; j < kFrameLengthsSize; ++j) {
     80     if (ValidRatesAndFrameLengths(8000, kFrameLengths[j])) {
     81       ASSERT_EQ(0, WebRtcVad_InitCore(self));
     82       EXPECT_EQ(0, WebRtcVad_CalculateFeatures(self, speech, kFrameLengths[j],
     83                                                features));
     84       for (int k = 0; k < kNumChannels; ++k) {
     85         EXPECT_EQ(kOffsetVector[k], features[k]);
     86       }
     87     }
     88   }
     89 
     90   free(self);
     91 }
     92 }  // namespace
     93