alexander | 3c79893 | 2021-03-26 21:42:19 +0000 | [diff] [blame] | 1 | /* |
| 2 | * Copyright (c) 2021 Arm Limited. All rights reserved. |
| 3 | * SPDX-License-Identifier: Apache-2.0 |
| 4 | * |
| 5 | * Licensed under the Apache License, Version 2.0 (the "License"); |
| 6 | * you may not use this file except in compliance with the License. |
| 7 | * You may obtain a copy of the License at |
| 8 | * |
| 9 | * http://www.apache.org/licenses/LICENSE-2.0 |
| 10 | * |
| 11 | * Unless required by applicable law or agreed to in writing, software |
| 12 | * distributed under the License is distributed on an "AS IS" BASIS, |
| 13 | * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
| 14 | * See the License for the specific language governing permissions and |
| 15 | * limitations under the License. |
| 16 | */ |
| 17 | #include "MelSpectrogram.hpp" |
| 18 | |
| 19 | #include "PlatformMath.hpp" |
alexander | 31ae9f0 | 2022-02-10 16:15:54 +0000 | [diff] [blame] | 20 | #include "log_macros.h" |
alexander | 3c79893 | 2021-03-26 21:42:19 +0000 | [diff] [blame] | 21 | |
| 22 | #include <cfloat> |
alexander | 31ae9f0 | 2022-02-10 16:15:54 +0000 | [diff] [blame] | 23 | #include <cinttypes> |
alexander | 3c79893 | 2021-03-26 21:42:19 +0000 | [diff] [blame] | 24 | |
| 25 | namespace arm { |
| 26 | namespace app { |
| 27 | namespace audio { |
| 28 | |
| 29 | MelSpecParams::MelSpecParams( |
| 30 | const float samplingFreq, |
| 31 | const uint32_t numFbankBins, |
| 32 | const float melLoFreq, |
| 33 | const float melHiFreq, |
| 34 | const uint32_t frameLen, |
| 35 | const bool useHtkMethod): |
| 36 | m_samplingFreq(samplingFreq), |
| 37 | m_numFbankBins(numFbankBins), |
| 38 | m_melLoFreq(melLoFreq), |
| 39 | m_melHiFreq(melHiFreq), |
| 40 | m_frameLen(frameLen), |
| 41 | |
| 42 | /* Smallest power of 2 >= frame length. */ |
| 43 | m_frameLenPadded(pow(2, ceil((log(frameLen)/log(2))))), |
| 44 | m_useHtkMethod(useHtkMethod) |
| 45 | {} |
| 46 | |
alexander | c350cdc | 2021-04-29 20:36:09 +0100 | [diff] [blame] | 47 | std::string MelSpecParams::Str() const |
alexander | 3c79893 | 2021-03-26 21:42:19 +0000 | [diff] [blame] | 48 | { |
| 49 | char strC[1024]; |
| 50 | snprintf(strC, sizeof(strC) - 1, "\n \ |
Kshitij Sisodia | f9c19ea | 2021-05-07 16:08:14 +0100 | [diff] [blame] | 51 | \n\t Sampling frequency: %f\ |
| 52 | \n\t Number of filter banks: %" PRIu32 "\ |
| 53 | \n\t Mel frequency limit (low): %f\ |
| 54 | \n\t Mel frequency limit (high): %f\ |
| 55 | \n\t Frame length: %" PRIu32 "\ |
| 56 | \n\t Padded frame length: %" PRIu32 "\ |
| 57 | \n\t Using HTK for Mel scale: %s\n", |
| 58 | this->m_samplingFreq, this->m_numFbankBins, this->m_melLoFreq, |
| 59 | this->m_melHiFreq, this->m_frameLen, |
| 60 | this->m_frameLenPadded, this->m_useHtkMethod ? "yes" : "no"); |
alexander | 3c79893 | 2021-03-26 21:42:19 +0000 | [diff] [blame] | 61 | return std::string{strC}; |
| 62 | } |
| 63 | |
| 64 | MelSpectrogram::MelSpectrogram(const MelSpecParams& params): |
Isabella Gottardi | 56ee620 | 2021-05-12 08:27:15 +0100 | [diff] [blame] | 65 | m_params(params), |
| 66 | m_filterBankInitialised(false) |
alexander | 3c79893 | 2021-03-26 21:42:19 +0000 | [diff] [blame] | 67 | { |
Isabella Gottardi | 56ee620 | 2021-05-12 08:27:15 +0100 | [diff] [blame] | 68 | this->m_buffer = std::vector<float>( |
| 69 | this->m_params.m_frameLenPadded, 0.0); |
| 70 | this->m_frame = std::vector<float>( |
| 71 | this->m_params.m_frameLenPadded, 0.0); |
| 72 | this->m_melEnergies = std::vector<float>( |
| 73 | this->m_params.m_numFbankBins, 0.0); |
alexander | 3c79893 | 2021-03-26 21:42:19 +0000 | [diff] [blame] | 74 | |
Isabella Gottardi | 56ee620 | 2021-05-12 08:27:15 +0100 | [diff] [blame] | 75 | this->m_windowFunc = std::vector<float>(this->m_params.m_frameLen); |
| 76 | const auto multiplier = static_cast<float>(2 * M_PI / this->m_params.m_frameLen); |
alexander | 3c79893 | 2021-03-26 21:42:19 +0000 | [diff] [blame] | 77 | |
| 78 | /* Create window function. */ |
Isabella Gottardi | 56ee620 | 2021-05-12 08:27:15 +0100 | [diff] [blame] | 79 | for (size_t i = 0; i < this->m_params.m_frameLen; ++i) { |
| 80 | this->m_windowFunc[i] = (0.5 - (0.5 * |
alexander | 3c79893 | 2021-03-26 21:42:19 +0000 | [diff] [blame] | 81 | math::MathUtils::CosineF32(static_cast<float>(i) * multiplier))); |
| 82 | } |
| 83 | |
Isabella Gottardi | 56ee620 | 2021-05-12 08:27:15 +0100 | [diff] [blame] | 84 | math::MathUtils::FftInitF32(this->m_params.m_frameLenPadded, this->m_fftInstance); |
| 85 | debug("Instantiated Mel Spectrogram object: %s\n", this->m_params.Str().c_str()); |
alexander | 3c79893 | 2021-03-26 21:42:19 +0000 | [diff] [blame] | 86 | } |
| 87 | |
| 88 | void MelSpectrogram::Init() |
| 89 | { |
alexander | c350cdc | 2021-04-29 20:36:09 +0100 | [diff] [blame] | 90 | this->InitMelFilterBank(); |
alexander | 3c79893 | 2021-03-26 21:42:19 +0000 | [diff] [blame] | 91 | } |
| 92 | |
| 93 | float MelSpectrogram::MelScale(const float freq, const bool useHTKMethod) |
| 94 | { |
| 95 | if (useHTKMethod) { |
| 96 | return 1127.0f * logf (1.0f + freq / 700.0f); |
| 97 | } else { |
| 98 | /* Slaney formula for mel scale. */ |
| 99 | float mel = freq / ms_freqStep; |
| 100 | |
| 101 | if (freq >= ms_minLogHz) { |
| 102 | mel = ms_minLogMel + logf(freq / ms_minLogHz) / ms_logStep; |
| 103 | } |
| 104 | return mel; |
| 105 | } |
| 106 | } |
| 107 | |
| 108 | float MelSpectrogram::InverseMelScale(const float melFreq, const bool useHTKMethod) |
| 109 | { |
| 110 | if (useHTKMethod) { |
| 111 | return 700.0f * (expf (melFreq / 1127.0f) - 1.0f); |
| 112 | } else { |
| 113 | /* Slaney formula for inverse mel scale. */ |
| 114 | float freq = ms_freqStep * melFreq; |
| 115 | |
| 116 | if (melFreq >= ms_minLogMel) { |
| 117 | freq = ms_minLogHz * expf(ms_logStep * (melFreq - ms_minLogMel)); |
| 118 | } |
| 119 | return freq; |
| 120 | } |
| 121 | } |
| 122 | |
| 123 | bool MelSpectrogram::ApplyMelFilterBank( |
| 124 | std::vector<float>& fftVec, |
| 125 | std::vector<std::vector<float>>& melFilterBank, |
alexander | c350cdc | 2021-04-29 20:36:09 +0100 | [diff] [blame] | 126 | std::vector<uint32_t>& filterBankFilterFirst, |
| 127 | std::vector<uint32_t>& filterBankFilterLast, |
alexander | 3c79893 | 2021-03-26 21:42:19 +0000 | [diff] [blame] | 128 | std::vector<float>& melEnergies) |
| 129 | { |
| 130 | const size_t numBanks = melEnergies.size(); |
| 131 | |
| 132 | if (numBanks != filterBankFilterFirst.size() || |
| 133 | numBanks != filterBankFilterLast.size()) { |
| 134 | printf_err("unexpected filter bank lengths\n"); |
| 135 | return false; |
| 136 | } |
| 137 | |
| 138 | for (size_t bin = 0; bin < numBanks; ++bin) { |
| 139 | auto filterBankIter = melFilterBank[bin].begin(); |
alexander | c350cdc | 2021-04-29 20:36:09 +0100 | [diff] [blame] | 140 | auto end = melFilterBank[bin].end(); |
alexander | 3c79893 | 2021-03-26 21:42:19 +0000 | [diff] [blame] | 141 | float melEnergy = FLT_MIN; /* Avoid log of zero at later stages */ |
alexander | c350cdc | 2021-04-29 20:36:09 +0100 | [diff] [blame] | 142 | const uint32_t firstIndex = filterBankFilterFirst[bin]; |
| 143 | const uint32_t lastIndex = std::min<int32_t>(filterBankFilterLast[bin], fftVec.size() - 1); |
alexander | 3c79893 | 2021-03-26 21:42:19 +0000 | [diff] [blame] | 144 | |
alexander | c350cdc | 2021-04-29 20:36:09 +0100 | [diff] [blame] | 145 | for (uint32_t i = firstIndex; i <= lastIndex && filterBankIter != end; ++i) { |
alexander | 3c79893 | 2021-03-26 21:42:19 +0000 | [diff] [blame] | 146 | float energyRep = math::MathUtils::SqrtF32(fftVec[i]); |
| 147 | melEnergy += (*filterBankIter++ * energyRep); |
| 148 | } |
| 149 | |
| 150 | melEnergies[bin] = melEnergy; |
| 151 | } |
| 152 | |
| 153 | return true; |
| 154 | } |
| 155 | |
| 156 | void MelSpectrogram::ConvertToLogarithmicScale(std::vector<float>& melEnergies) |
| 157 | { |
alexander | c350cdc | 2021-04-29 20:36:09 +0100 | [diff] [blame] | 158 | for (float& melEnergy : melEnergies) { |
| 159 | melEnergy = logf(melEnergy); |
alexander | 3c79893 | 2021-03-26 21:42:19 +0000 | [diff] [blame] | 160 | } |
| 161 | } |
| 162 | |
alexander | c350cdc | 2021-04-29 20:36:09 +0100 | [diff] [blame] | 163 | void MelSpectrogram::ConvertToPowerSpectrum() |
alexander | 3c79893 | 2021-03-26 21:42:19 +0000 | [diff] [blame] | 164 | { |
Isabella Gottardi | 56ee620 | 2021-05-12 08:27:15 +0100 | [diff] [blame] | 165 | const uint32_t halfDim = this->m_buffer.size() / 2; |
alexander | 3c79893 | 2021-03-26 21:42:19 +0000 | [diff] [blame] | 166 | |
| 167 | /* Handle this special case. */ |
Isabella Gottardi | 56ee620 | 2021-05-12 08:27:15 +0100 | [diff] [blame] | 168 | float firstEnergy = this->m_buffer[0] * this->m_buffer[0]; |
| 169 | float lastEnergy = this->m_buffer[1] * this->m_buffer[1]; |
alexander | 3c79893 | 2021-03-26 21:42:19 +0000 | [diff] [blame] | 170 | |
| 171 | math::MathUtils::ComplexMagnitudeSquaredF32( |
Isabella Gottardi | 56ee620 | 2021-05-12 08:27:15 +0100 | [diff] [blame] | 172 | this->m_buffer.data(), |
| 173 | this->m_buffer.size(), |
| 174 | this->m_buffer.data(), |
| 175 | this->m_buffer.size()/2); |
alexander | 3c79893 | 2021-03-26 21:42:19 +0000 | [diff] [blame] | 176 | |
Isabella Gottardi | 56ee620 | 2021-05-12 08:27:15 +0100 | [diff] [blame] | 177 | this->m_buffer[0] = firstEnergy; |
| 178 | this->m_buffer[halfDim] = lastEnergy; |
alexander | 3c79893 | 2021-03-26 21:42:19 +0000 | [diff] [blame] | 179 | } |
| 180 | |
| 181 | float MelSpectrogram::GetMelFilterBankNormaliser( |
| 182 | const float& leftMel, |
| 183 | const float& rightMel, |
| 184 | const bool useHTKMethod) |
| 185 | { |
| 186 | UNUSED(leftMel); |
| 187 | UNUSED(rightMel); |
| 188 | UNUSED(useHTKMethod); |
| 189 | |
| 190 | /* By default, no normalisation => return 1 */ |
| 191 | return 1.f; |
| 192 | } |
| 193 | |
alexander | c350cdc | 2021-04-29 20:36:09 +0100 | [diff] [blame] | 194 | void MelSpectrogram::InitMelFilterBank() |
alexander | 3c79893 | 2021-03-26 21:42:19 +0000 | [diff] [blame] | 195 | { |
alexander | c350cdc | 2021-04-29 20:36:09 +0100 | [diff] [blame] | 196 | if (!this->IsMelFilterBankInited()) { |
Isabella Gottardi | 56ee620 | 2021-05-12 08:27:15 +0100 | [diff] [blame] | 197 | this->m_melFilterBank = this->CreateMelFilterBank(); |
| 198 | this->m_filterBankInitialised = true; |
alexander | 3c79893 | 2021-03-26 21:42:19 +0000 | [diff] [blame] | 199 | } |
| 200 | } |
| 201 | |
alexander | c350cdc | 2021-04-29 20:36:09 +0100 | [diff] [blame] | 202 | bool MelSpectrogram::IsMelFilterBankInited() const |
alexander | 3c79893 | 2021-03-26 21:42:19 +0000 | [diff] [blame] | 203 | { |
Isabella Gottardi | 56ee620 | 2021-05-12 08:27:15 +0100 | [diff] [blame] | 204 | return this->m_filterBankInitialised; |
alexander | 3c79893 | 2021-03-26 21:42:19 +0000 | [diff] [blame] | 205 | } |
| 206 | |
| 207 | std::vector<float> MelSpectrogram::ComputeMelSpec(const std::vector<int16_t>& audioData, float trainingMean) |
| 208 | { |
alexander | c350cdc | 2021-04-29 20:36:09 +0100 | [diff] [blame] | 209 | this->InitMelFilterBank(); |
alexander | 3c79893 | 2021-03-26 21:42:19 +0000 | [diff] [blame] | 210 | |
| 211 | /* TensorFlow way of normalizing .wav data to (-1, 1). */ |
| 212 | constexpr float normaliser = 1.0/(1<<15); |
Isabella Gottardi | 56ee620 | 2021-05-12 08:27:15 +0100 | [diff] [blame] | 213 | for (size_t i = 0; i < this->m_params.m_frameLen; ++i) { |
| 214 | this->m_frame[i] = static_cast<float>(audioData[i]) * normaliser; |
alexander | 3c79893 | 2021-03-26 21:42:19 +0000 | [diff] [blame] | 215 | } |
| 216 | |
| 217 | /* Apply window function to input frame. */ |
Isabella Gottardi | 56ee620 | 2021-05-12 08:27:15 +0100 | [diff] [blame] | 218 | for(size_t i = 0; i < this->m_params.m_frameLen; ++i) { |
| 219 | this->m_frame[i] *= this->m_windowFunc[i]; |
alexander | 3c79893 | 2021-03-26 21:42:19 +0000 | [diff] [blame] | 220 | } |
| 221 | |
| 222 | /* Set remaining frame values to 0. */ |
Isabella Gottardi | 56ee620 | 2021-05-12 08:27:15 +0100 | [diff] [blame] | 223 | std::fill(this->m_frame.begin() + this->m_params.m_frameLen,this->m_frame.end(), 0); |
alexander | 3c79893 | 2021-03-26 21:42:19 +0000 | [diff] [blame] | 224 | |
| 225 | /* Compute FFT. */ |
Isabella Gottardi | 56ee620 | 2021-05-12 08:27:15 +0100 | [diff] [blame] | 226 | math::MathUtils::FftF32(this->m_frame, this->m_buffer, this->m_fftInstance); |
alexander | 3c79893 | 2021-03-26 21:42:19 +0000 | [diff] [blame] | 227 | |
| 228 | /* Convert to power spectrum. */ |
alexander | c350cdc | 2021-04-29 20:36:09 +0100 | [diff] [blame] | 229 | this->ConvertToPowerSpectrum(); |
alexander | 3c79893 | 2021-03-26 21:42:19 +0000 | [diff] [blame] | 230 | |
| 231 | /* Apply mel filterbanks. */ |
Isabella Gottardi | 56ee620 | 2021-05-12 08:27:15 +0100 | [diff] [blame] | 232 | if (!this->ApplyMelFilterBank(this->m_buffer, |
| 233 | this->m_melFilterBank, |
| 234 | this->m_filterBankFilterFirst, |
| 235 | this->m_filterBankFilterLast, |
| 236 | this->m_melEnergies)) { |
alexander | 3c79893 | 2021-03-26 21:42:19 +0000 | [diff] [blame] | 237 | printf_err("Failed to apply MEL filter banks\n"); |
| 238 | } |
| 239 | |
| 240 | /* Convert to logarithmic scale */ |
Isabella Gottardi | 56ee620 | 2021-05-12 08:27:15 +0100 | [diff] [blame] | 241 | this->ConvertToLogarithmicScale(this->m_melEnergies); |
alexander | 3c79893 | 2021-03-26 21:42:19 +0000 | [diff] [blame] | 242 | |
| 243 | /* Perform mean subtraction. */ |
Isabella Gottardi | 56ee620 | 2021-05-12 08:27:15 +0100 | [diff] [blame] | 244 | for (auto& energy:this->m_melEnergies) { |
alexander | 3c79893 | 2021-03-26 21:42:19 +0000 | [diff] [blame] | 245 | energy -= trainingMean; |
| 246 | } |
| 247 | |
Isabella Gottardi | 56ee620 | 2021-05-12 08:27:15 +0100 | [diff] [blame] | 248 | return this->m_melEnergies; |
alexander | 3c79893 | 2021-03-26 21:42:19 +0000 | [diff] [blame] | 249 | } |
| 250 | |
alexander | c350cdc | 2021-04-29 20:36:09 +0100 | [diff] [blame] | 251 | std::vector<std::vector<float>> MelSpectrogram::CreateMelFilterBank() |
alexander | 3c79893 | 2021-03-26 21:42:19 +0000 | [diff] [blame] | 252 | { |
Isabella Gottardi | 56ee620 | 2021-05-12 08:27:15 +0100 | [diff] [blame] | 253 | size_t numFftBins = this->m_params.m_frameLenPadded / 2; |
| 254 | float fftBinWidth = static_cast<float>(this->m_params.m_samplingFreq) / this->m_params.m_frameLenPadded; |
alexander | 3c79893 | 2021-03-26 21:42:19 +0000 | [diff] [blame] | 255 | |
Isabella Gottardi | 56ee620 | 2021-05-12 08:27:15 +0100 | [diff] [blame] | 256 | float melLowFreq = MelSpectrogram::MelScale(this->m_params.m_melLoFreq, |
| 257 | this->m_params.m_useHtkMethod); |
| 258 | float melHighFreq = MelSpectrogram::MelScale(this->m_params.m_melHiFreq, |
| 259 | this->m_params.m_useHtkMethod); |
| 260 | float melFreqDelta = (melHighFreq - melLowFreq) / (this->m_params.m_numFbankBins + 1); |
alexander | 3c79893 | 2021-03-26 21:42:19 +0000 | [diff] [blame] | 261 | |
| 262 | std::vector<float> thisBin = std::vector<float>(numFftBins); |
| 263 | std::vector<std::vector<float>> melFilterBank( |
Isabella Gottardi | 56ee620 | 2021-05-12 08:27:15 +0100 | [diff] [blame] | 264 | this->m_params.m_numFbankBins); |
| 265 | this->m_filterBankFilterFirst = |
| 266 | std::vector<uint32_t>(this->m_params.m_numFbankBins); |
| 267 | this->m_filterBankFilterLast = |
| 268 | std::vector<uint32_t>(this->m_params.m_numFbankBins); |
alexander | 3c79893 | 2021-03-26 21:42:19 +0000 | [diff] [blame] | 269 | |
Isabella Gottardi | 56ee620 | 2021-05-12 08:27:15 +0100 | [diff] [blame] | 270 | for (size_t bin = 0; bin < this->m_params.m_numFbankBins; bin++) { |
alexander | 3c79893 | 2021-03-26 21:42:19 +0000 | [diff] [blame] | 271 | float leftMel = melLowFreq + bin * melFreqDelta; |
| 272 | float centerMel = melLowFreq + (bin + 1) * melFreqDelta; |
| 273 | float rightMel = melLowFreq + (bin + 2) * melFreqDelta; |
| 274 | |
alexander | c350cdc | 2021-04-29 20:36:09 +0100 | [diff] [blame] | 275 | uint32_t firstIndex = 0; |
| 276 | uint32_t lastIndex = 0; |
| 277 | bool firstIndexFound = false; |
Isabella Gottardi | 56ee620 | 2021-05-12 08:27:15 +0100 | [diff] [blame] | 278 | const float normaliser = this->GetMelFilterBankNormaliser(leftMel, rightMel, this->m_params.m_useHtkMethod); |
alexander | 3c79893 | 2021-03-26 21:42:19 +0000 | [diff] [blame] | 279 | |
| 280 | for (size_t i = 0; i < numFftBins; ++i) { |
| 281 | float freq = (fftBinWidth * i); /* Center freq of this fft bin. */ |
Isabella Gottardi | 56ee620 | 2021-05-12 08:27:15 +0100 | [diff] [blame] | 282 | float mel = MelSpectrogram::MelScale(freq, this->m_params.m_useHtkMethod); |
alexander | 3c79893 | 2021-03-26 21:42:19 +0000 | [diff] [blame] | 283 | thisBin[i] = 0.0; |
| 284 | |
| 285 | if (mel > leftMel && mel < rightMel) { |
| 286 | float weight; |
| 287 | if (mel <= centerMel) { |
| 288 | weight = (mel - leftMel) / (centerMel - leftMel); |
| 289 | } else { |
| 290 | weight = (rightMel - mel) / (rightMel - centerMel); |
| 291 | } |
| 292 | |
| 293 | thisBin[i] = weight * normaliser; |
alexander | c350cdc | 2021-04-29 20:36:09 +0100 | [diff] [blame] | 294 | if (!firstIndexFound) { |
alexander | 3c79893 | 2021-03-26 21:42:19 +0000 | [diff] [blame] | 295 | firstIndex = i; |
alexander | c350cdc | 2021-04-29 20:36:09 +0100 | [diff] [blame] | 296 | firstIndexFound = true; |
alexander | 3c79893 | 2021-03-26 21:42:19 +0000 | [diff] [blame] | 297 | } |
| 298 | lastIndex = i; |
| 299 | } |
| 300 | } |
| 301 | |
Isabella Gottardi | 56ee620 | 2021-05-12 08:27:15 +0100 | [diff] [blame] | 302 | this->m_filterBankFilterFirst[bin] = firstIndex; |
| 303 | this->m_filterBankFilterLast[bin] = lastIndex; |
alexander | 3c79893 | 2021-03-26 21:42:19 +0000 | [diff] [blame] | 304 | |
| 305 | /* Copy the part we care about. */ |
alexander | c350cdc | 2021-04-29 20:36:09 +0100 | [diff] [blame] | 306 | for (uint32_t i = firstIndex; i <= lastIndex; ++i) { |
alexander | 3c79893 | 2021-03-26 21:42:19 +0000 | [diff] [blame] | 307 | melFilterBank[bin].push_back(thisBin[i]); |
| 308 | } |
| 309 | } |
| 310 | |
| 311 | return melFilterBank; |
| 312 | } |
| 313 | |
| 314 | } /* namespace audio */ |
| 315 | } /* namespace app */ |
| 316 | } /* namespace arm */ |