/src/mozilla-central/dom/media/webaudio/FFTBlock.cpp
Line | Count | Source (jump to first uncovered line) |
1 | | /* -*- Mode: C++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4 -*- */ |
2 | | /* vim:set ts=4 sw=4 sts=4 et cindent: */ |
3 | | /* |
4 | | * Copyright (C) 2010 Google Inc. All rights reserved. |
5 | | * |
6 | | * Redistribution and use in source and binary forms, with or without |
7 | | * modification, are permitted provided that the following conditions |
8 | | * are met: |
9 | | * |
10 | | * 1. Redistributions of source code must retain the above copyright |
11 | | * notice, this list of conditions and the following disclaimer. |
12 | | * 2. Redistributions in binary form must reproduce the above copyright |
13 | | * notice, this list of conditions and the following disclaimer in the |
14 | | * documentation and/or other materials provided with the distribution. |
15 | | * 3. Neither the name of Apple Computer, Inc. ("Apple") nor the names of |
16 | | * its contributors may be used to endorse or promote products derived |
17 | | * from this software without specific prior written permission. |
18 | | * |
19 | | * THIS SOFTWARE IS PROVIDED BY APPLE AND ITS CONTRIBUTORS "AS IS" AND ANY |
20 | | * EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED |
21 | | * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE |
22 | | * DISCLAIMED. IN NO EVENT SHALL APPLE OR ITS CONTRIBUTORS BE LIABLE FOR ANY |
23 | | * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES |
24 | | * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; |
25 | | * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND |
26 | | * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT |
27 | | * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF |
28 | | * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. |
29 | | */ |
30 | | |
31 | | #include "FFTBlock.h" |
32 | | |
33 | | #include <complex> |
34 | | |
35 | | namespace mozilla { |
36 | | |
37 | | typedef std::complex<double> Complex; |
38 | | |
39 | | #ifdef MOZ_LIBAV_FFT |
40 | | FFmpegRDFTFuncs FFTBlock::sRDFTFuncs; |
41 | | #endif |
42 | | |
43 | | FFTBlock* FFTBlock::CreateInterpolatedBlock(const FFTBlock& block0, const FFTBlock& block1, double interp) |
44 | 0 | { |
45 | 0 | FFTBlock* newBlock = new FFTBlock(block0.FFTSize()); |
46 | 0 |
|
47 | 0 | newBlock->InterpolateFrequencyComponents(block0, block1, interp); |
48 | 0 |
|
49 | 0 | // In the time-domain, the 2nd half of the response must be zero, to avoid circular convolution aliasing... |
50 | 0 | int fftSize = newBlock->FFTSize(); |
51 | 0 | AlignedTArray<float> buffer(fftSize); |
52 | 0 | newBlock->GetInverseWithoutScaling(buffer.Elements()); |
53 | 0 | AudioBufferInPlaceScale(buffer.Elements(), 1.0f / fftSize, fftSize / 2); |
54 | 0 | PodZero(buffer.Elements() + fftSize / 2, fftSize / 2); |
55 | 0 |
|
56 | 0 | // Put back into frequency domain. |
57 | 0 | newBlock->PerformFFT(buffer.Elements()); |
58 | 0 |
|
59 | 0 | return newBlock; |
60 | 0 | } |
61 | | |
62 | | void FFTBlock::InterpolateFrequencyComponents(const FFTBlock& block0, const FFTBlock& block1, double interp) |
63 | 0 | { |
64 | 0 | // FIXME : with some work, this method could be optimized |
65 | 0 |
|
66 | 0 | ComplexU* dft = mOutputBuffer.Elements(); |
67 | 0 |
|
68 | 0 | const ComplexU* dft1 = block0.mOutputBuffer.Elements(); |
69 | 0 | const ComplexU* dft2 = block1.mOutputBuffer.Elements(); |
70 | 0 |
|
71 | 0 | MOZ_ASSERT(mFFTSize == block0.FFTSize()); |
72 | 0 | MOZ_ASSERT(mFFTSize == block1.FFTSize()); |
73 | 0 | double s1base = (1.0 - interp); |
74 | 0 | double s2base = interp; |
75 | 0 |
|
76 | 0 | double phaseAccum = 0.0; |
77 | 0 | double lastPhase1 = 0.0; |
78 | 0 | double lastPhase2 = 0.0; |
79 | 0 |
|
80 | 0 | int n = mFFTSize / 2; |
81 | 0 |
|
82 | 0 | dft[0].r = static_cast<float>(s1base * dft1[0].r + s2base * dft2[0].r); |
83 | 0 | dft[n].r = static_cast<float>(s1base * dft1[n].r + s2base * dft2[n].r); |
84 | 0 |
|
85 | 0 | for (int i = 1; i < n; ++i) { |
86 | 0 | Complex c1(dft1[i].r, dft1[i].i); |
87 | 0 | Complex c2(dft2[i].r, dft2[i].i); |
88 | 0 |
|
89 | 0 | double mag1 = abs(c1); |
90 | 0 | double mag2 = abs(c2); |
91 | 0 |
|
92 | 0 | // Interpolate magnitudes in decibels |
93 | 0 | double mag1db = 20.0 * log10(mag1); |
94 | 0 | double mag2db = 20.0 * log10(mag2); |
95 | 0 |
|
96 | 0 | double s1 = s1base; |
97 | 0 | double s2 = s2base; |
98 | 0 |
|
99 | 0 | double magdbdiff = mag1db - mag2db; |
100 | 0 |
|
101 | 0 | // Empirical tweak to retain higher-frequency zeroes |
102 | 0 | double threshold = (i > 16) ? 5.0 : 2.0; |
103 | 0 |
|
104 | 0 | if (magdbdiff < -threshold && mag1db < 0.0) { |
105 | 0 | s1 = pow(s1, 0.75); |
106 | 0 | s2 = 1.0 - s1; |
107 | 0 | } else if (magdbdiff > threshold && mag2db < 0.0) { |
108 | 0 | s2 = pow(s2, 0.75); |
109 | 0 | s1 = 1.0 - s2; |
110 | 0 | } |
111 | 0 |
|
112 | 0 | // Average magnitude by decibels instead of linearly |
113 | 0 | double magdb = s1 * mag1db + s2 * mag2db; |
114 | 0 | double mag = pow(10.0, 0.05 * magdb); |
115 | 0 |
|
116 | 0 | // Now, deal with phase |
117 | 0 | double phase1 = arg(c1); |
118 | 0 | double phase2 = arg(c2); |
119 | 0 |
|
120 | 0 | double deltaPhase1 = phase1 - lastPhase1; |
121 | 0 | double deltaPhase2 = phase2 - lastPhase2; |
122 | 0 | lastPhase1 = phase1; |
123 | 0 | lastPhase2 = phase2; |
124 | 0 |
|
125 | 0 | // Unwrap phase deltas |
126 | 0 | if (deltaPhase1 > M_PI) |
127 | 0 | deltaPhase1 -= 2.0 * M_PI; |
128 | 0 | if (deltaPhase1 < -M_PI) |
129 | 0 | deltaPhase1 += 2.0 * M_PI; |
130 | 0 | if (deltaPhase2 > M_PI) |
131 | 0 | deltaPhase2 -= 2.0 * M_PI; |
132 | 0 | if (deltaPhase2 < -M_PI) |
133 | 0 | deltaPhase2 += 2.0 * M_PI; |
134 | 0 |
|
135 | 0 | // Blend group-delays |
136 | 0 | double deltaPhaseBlend; |
137 | 0 |
|
138 | 0 | if (deltaPhase1 - deltaPhase2 > M_PI) |
139 | 0 | deltaPhaseBlend = s1 * deltaPhase1 + s2 * (2.0 * M_PI + deltaPhase2); |
140 | 0 | else if (deltaPhase2 - deltaPhase1 > M_PI) |
141 | 0 | deltaPhaseBlend = s1 * (2.0 * M_PI + deltaPhase1) + s2 * deltaPhase2; |
142 | 0 | else |
143 | 0 | deltaPhaseBlend = s1 * deltaPhase1 + s2 * deltaPhase2; |
144 | 0 |
|
145 | 0 | phaseAccum += deltaPhaseBlend; |
146 | 0 |
|
147 | 0 | // Unwrap |
148 | 0 | if (phaseAccum > M_PI) |
149 | 0 | phaseAccum -= 2.0 * M_PI; |
150 | 0 | if (phaseAccum < -M_PI) |
151 | 0 | phaseAccum += 2.0 * M_PI; |
152 | 0 |
|
153 | 0 | dft[i].r = static_cast<float>(mag * cos(phaseAccum)); |
154 | 0 | dft[i].i = static_cast<float>(mag * sin(phaseAccum)); |
155 | 0 | } |
156 | 0 | } |
157 | | |
158 | | double FFTBlock::ExtractAverageGroupDelay() |
159 | 0 | { |
160 | 0 | ComplexU* dft = mOutputBuffer.Elements(); |
161 | 0 |
|
162 | 0 | double aveSum = 0.0; |
163 | 0 | double weightSum = 0.0; |
164 | 0 | double lastPhase = 0.0; |
165 | 0 |
|
166 | 0 | int halfSize = FFTSize() / 2; |
167 | 0 |
|
168 | 0 | const double kSamplePhaseDelay = (2.0 * M_PI) / double(FFTSize()); |
169 | 0 |
|
170 | 0 | // Remove DC offset |
171 | 0 | dft[0].r = 0.0f; |
172 | 0 |
|
173 | 0 | // Calculate weighted average group delay |
174 | 0 | for (int i = 1; i < halfSize; i++) { |
175 | 0 | Complex c(dft[i].r, dft[i].i); |
176 | 0 | double mag = abs(c); |
177 | 0 | double phase = arg(c); |
178 | 0 |
|
179 | 0 | double deltaPhase = phase - lastPhase; |
180 | 0 | lastPhase = phase; |
181 | 0 |
|
182 | 0 | // Unwrap |
183 | 0 | if (deltaPhase < -M_PI) |
184 | 0 | deltaPhase += 2.0 * M_PI; |
185 | 0 | if (deltaPhase > M_PI) |
186 | 0 | deltaPhase -= 2.0 * M_PI; |
187 | 0 |
|
188 | 0 | aveSum += mag * deltaPhase; |
189 | 0 | weightSum += mag; |
190 | 0 | } |
191 | 0 |
|
192 | 0 | // Note how we invert the phase delta wrt frequency since this is how group delay is defined |
193 | 0 | double ave = aveSum / weightSum; |
194 | 0 | double aveSampleDelay = -ave / kSamplePhaseDelay; |
195 | 0 |
|
196 | 0 | // Leave 20 sample headroom (for leading edge of impulse) |
197 | 0 | aveSampleDelay -= 20.0; |
198 | 0 | if (aveSampleDelay <= 0.0) |
199 | 0 | return 0.0; |
200 | 0 | |
201 | 0 | // Remove average group delay (minus 20 samples for headroom) |
202 | 0 | AddConstantGroupDelay(-aveSampleDelay); |
203 | 0 |
|
204 | 0 | return aveSampleDelay; |
205 | 0 | } |
206 | | |
207 | | void FFTBlock::AddConstantGroupDelay(double sampleFrameDelay) |
208 | 0 | { |
209 | 0 | int halfSize = FFTSize() / 2; |
210 | 0 |
|
211 | 0 | ComplexU* dft = mOutputBuffer.Elements(); |
212 | 0 |
|
213 | 0 | const double kSamplePhaseDelay = (2.0 * M_PI) / double(FFTSize()); |
214 | 0 |
|
215 | 0 | double phaseAdj = -sampleFrameDelay * kSamplePhaseDelay; |
216 | 0 |
|
217 | 0 | // Add constant group delay |
218 | 0 | for (int i = 1; i < halfSize; i++) { |
219 | 0 | Complex c(dft[i].r, dft[i].i); |
220 | 0 | double mag = abs(c); |
221 | 0 | double phase = arg(c); |
222 | 0 |
|
223 | 0 | phase += i * phaseAdj; |
224 | 0 |
|
225 | 0 | dft[i].r = static_cast<float>(mag * cos(phase)); |
226 | 0 | dft[i].i = static_cast<float>(mag * sin(phase)); |
227 | 0 | } |
228 | 0 | } |
229 | | |
230 | | } // namespace mozilla |