-
Notifications
You must be signed in to change notification settings - Fork 1.8k
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
Merge pull request #1640 from zh794390558/frontend
[speechx] Frontend refactor
- Loading branch information
Showing
26 changed files
with
210 additions
and
178 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -1,10 +1,2 @@ | ||
| project(frontend) | ||
|
|
||
| add_library(frontend STATIC | ||
| normalizer.cc | ||
| linear_spectrogram.cc | ||
| audio_cache.cc | ||
| feature_cache.cc | ||
| ) | ||
|
|
||
| target_link_libraries(frontend PUBLIC kaldi-matrix) | ||
| add_subdirectory(audio) |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,11 @@ | ||
| project(frontend) | ||
|
|
||
| add_library(frontend STATIC | ||
| cmvn.cc | ||
| db_norm.cc | ||
| linear_spectrogram.cc | ||
| audio_cache.cc | ||
| feature_cache.cc | ||
| ) | ||
|
|
||
| target_link_libraries(frontend PUBLIC kaldi-matrix) |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,48 @@ | ||
| // Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved. | ||
| // | ||
| // Licensed under the Apache License, Version 2.0 (the "License"); | ||
| // you may not use this file except in compliance with the License. | ||
| // You may obtain a copy of the License at | ||
| // | ||
| // http://www.apache.org/licenses/LICENSE-2.0 | ||
| // | ||
| // Unless required by applicable law or agreed to in writing, software | ||
| // distributed under the License is distributed on an "AS IS" BASIS, | ||
| // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. | ||
| // See the License for the specific language governing permissions and | ||
| // limitations under the License. | ||
|
|
||
| #pragma once | ||
|
|
||
| #include "base/common.h" | ||
| #include "frontend/audio/frontend_itf.h" | ||
| #include "kaldi/matrix/kaldi-matrix.h" | ||
| #include "kaldi/util/options-itf.h" | ||
|
|
||
| namespace ppspeech { | ||
|
|
||
| class CMVN : public FrontendInterface { | ||
| public: | ||
| explicit CMVN(std::string cmvn_file, | ||
| std::unique_ptr<FrontendInterface> base_extractor); | ||
| virtual void Accept(const kaldi::VectorBase<kaldi::BaseFloat>& inputs); | ||
|
|
||
| // the length of feats = feature_row * feature_dim, | ||
| // the Matrix is squashed into Vector | ||
| virtual bool Read(kaldi::Vector<kaldi::BaseFloat>* feats); | ||
| // the dim_ is the feautre dim. | ||
| virtual size_t Dim() const { return dim_; } | ||
| virtual void SetFinished() { base_extractor_->SetFinished(); } | ||
| virtual bool IsFinished() const { return base_extractor_->IsFinished(); } | ||
| virtual void Reset() { base_extractor_->Reset(); } | ||
|
|
||
| private: | ||
| void Compute(kaldi::VectorBase<kaldi::BaseFloat>* feats) const; | ||
| void ApplyCMVN(kaldi::MatrixBase<BaseFloat>* feats); | ||
| kaldi::Matrix<double> stats_; | ||
| std::unique_ptr<FrontendInterface> base_extractor_; | ||
| size_t dim_; | ||
| bool var_norm_; | ||
| }; | ||
|
|
||
| } // namespace ppspeech |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,95 @@ | ||
| // Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved. | ||
| // | ||
| // Licensed under the Apache License, Version 2.0 (the "License"); | ||
| // you may not use this file except in compliance with the License. | ||
| // You may obtain a copy of the License at | ||
| // | ||
| // http://www.apache.org/licenses/LICENSE-2.0 | ||
| // | ||
| // Unless required by applicable law or agreed to in writing, software | ||
| // distributed under the License is distributed on an "AS IS" BASIS, | ||
| // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. | ||
| // See the License for the specific language governing permissions and | ||
| // limitations under the License. | ||
|
|
||
|
|
||
| #include "frontend/audio/db_norm.h" | ||
| #include "kaldi/feat/cmvn.h" | ||
| #include "kaldi/util/kaldi-io.h" | ||
|
|
||
| namespace ppspeech { | ||
|
|
||
| using kaldi::Vector; | ||
| using kaldi::VectorBase; | ||
| using kaldi::BaseFloat; | ||
| using std::vector; | ||
| using kaldi::SubVector; | ||
| using std::unique_ptr; | ||
|
|
||
| DecibelNormalizer::DecibelNormalizer( | ||
| const DecibelNormalizerOptions& opts, | ||
| std::unique_ptr<FrontendInterface> base_extractor) { | ||
| base_extractor_ = std::move(base_extractor); | ||
| opts_ = opts; | ||
| dim_ = 1; | ||
| } | ||
|
|
||
| void DecibelNormalizer::Accept(const kaldi::VectorBase<BaseFloat>& waves) { | ||
| base_extractor_->Accept(waves); | ||
| } | ||
|
|
||
| bool DecibelNormalizer::Read(kaldi::Vector<BaseFloat>* waves) { | ||
| if (base_extractor_->Read(waves) == false || waves->Dim() == 0) { | ||
| return false; | ||
| } | ||
| Compute(waves); | ||
| return true; | ||
| } | ||
|
|
||
| bool DecibelNormalizer::Compute(VectorBase<BaseFloat>* waves) const { | ||
| // calculate db rms | ||
| BaseFloat rms_db = 0.0; | ||
| BaseFloat mean_square = 0.0; | ||
| BaseFloat gain = 0.0; | ||
| BaseFloat wave_float_normlization = 1.0f / (std::pow(2, 16 - 1)); | ||
|
|
||
| vector<BaseFloat> samples; | ||
| samples.resize(waves->Dim()); | ||
| for (size_t i = 0; i < samples.size(); ++i) { | ||
| samples[i] = (*waves)(i); | ||
| } | ||
|
|
||
| // square | ||
| for (auto& d : samples) { | ||
| if (opts_.convert_int_float) { | ||
| d = d * wave_float_normlization; | ||
| } | ||
| mean_square += d * d; | ||
| } | ||
|
|
||
| // mean | ||
| mean_square /= samples.size(); | ||
| rms_db = 10 * std::log10(mean_square); | ||
| gain = opts_.target_db - rms_db; | ||
|
|
||
| if (gain > opts_.max_gain_db) { | ||
| LOG(ERROR) | ||
| << "Unable to normalize segment to " << opts_.target_db << "dB," | ||
| << "because the the probable gain have exceeds opts_.max_gain_db" | ||
| << opts_.max_gain_db << "dB."; | ||
| return false; | ||
| } | ||
|
|
||
| // Note that this is an in-place transformation. | ||
| for (auto& item : samples) { | ||
| // python item *= 10.0 ** (gain / 20.0) | ||
| item *= std::pow(10.0, gain / 20.0); | ||
| } | ||
|
|
||
| std::memcpy( | ||
| waves->Data(), samples.data(), sizeof(BaseFloat) * samples.size()); | ||
| return true; | ||
| } | ||
|
|
||
|
|
||
| } // namespace ppspeech |
Oops, something went wrong.