tom@419: // Copyright 2010, Google
tom@419: //
tom@419: // AIM-C: A C++ implementation of the Auditory Image Model
tom@419: // http://www.acousticscale.org/AIMC
tom@419: //
tom@419: // Licensed under the Apache License, Version 2.0 (the "License");
tom@419: // you may not use this file except in compliance with the License.
tom@419: // You may obtain a copy of the License at
tom@419: //
tom@419: //     http://www.apache.org/licenses/LICENSE-2.0
tom@419: //
tom@419: // Unless required by applicable law or agreed to in writing, software
tom@419: // distributed under the License is distributed on an "AS IS" BASIS,
tom@419: // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
tom@419: // See the License for the specific language governing permissions and
tom@419: // limitations under the License.
tom@419: 
tom@419: /*!
tom@419:  * \author Thomas Walters <tom@acousticscale.org>
tom@419:  * \date created 2010-06-14
tom@419:  * \version \$Id$
tom@419:  */
tom@419: 
tom@419: #include "Modules/Features/ModuleBoxes.h"
tom@419: 
tom@419: namespace aimc {
tom@419: ModuleBoxes::ModuleBoxes(Parameters *params) : Module(params) {
tom@419:   module_description_ = "'Box-cutting' routine to generate dense features";
tom@419:   module_identifier_ = "boxes";
tom@419:   module_type_ = "features";
tom@419:   module_version_ = "$Id$";
tom@419: 
tom@419:   box_size_spectral_ = parameters_->DefaultInt("boxes.spectral_size", 16);
tom@419:   box_size_temporal_ = parameters_->DefaultInt("boxes.temporal_size", 32);
tom@419: }
tom@419: 
tom@419: ModuleBoxes::~ModuleBoxes() {
tom@419: }
tom@419: 
tom@419: bool ModuleBoxes::InitializeInternal(const SignalBank &input) {
tom@419:   // Copy the parameters of the input signal bank into internal variables, so
tom@419:   // that they can be checked later.
tom@419:   sample_rate_ = input.sample_rate();
tom@419:   buffer_length_ = input.buffer_length();
tom@419:   channel_count_ = input.channel_count();
tom@419: 
tom@419:   int channels_height = box_size_spectral_;
tom@419:   while (channels_height < channel_count_ / 2) {
tom@419:     int top = channel_count_ - 1;
tom@419:     while (top - channels_height >= 0) {
tom@419:       box_limits_channels_.push_back(std::make_pair(top,
tom@419:                                                     top - channels_height));
tom@419:       LOG_INFO("ch: t %d, b %d", top, top - channels_height);
tom@419:       top -= channels_height / 2;
tom@419:     }
tom@419:     channels_height *= 2;
tom@419:   }
tom@419: 
tom@419:   int temporal_width = box_size_temporal_;
tom@419:   while (temporal_width < buffer_length_) {
tom@419:     box_limits_time_.push_back(temporal_width);
tom@419:     LOG_INFO("sp: %d", temporal_width);
tom@419:     temporal_width *= 2;
tom@419:   }
tom@419: 
tom@419:   box_count_ = box_limits_time_.size() * box_limits_channels_.size();
tom@419:   feature_size_ = box_size_spectral_ + box_size_temporal_;
tom@419:   LOG_INFO("Total box count is %d", box_count_);
tom@419:   LOG_INFO("Total feature size is %d", feature_size_);
tom@419: 
tom@419:   output_.Initialize(box_count_, feature_size_, 1.0f);
tom@419:   return true;
tom@419: }
tom@419: 
tom@419: void ModuleBoxes::ResetInternal() {
tom@419: }
tom@419: 
tom@419: void ModuleBoxes::Process(const SignalBank &input) {
tom@419:   // Check to see if the module has been initialized. If not, processing
tom@419:   // should not continue.
tom@419:   if (!initialized_) {
tom@419:     LOG_ERROR(_T("Module %s not initialized."), module_identifier_.c_str());
tom@419:     return;
tom@419:   }
tom@419: 
tom@419:   // Check that ths input this time is the same as the input passed to
tom@419:   // Initialize()
tom@419:   if (buffer_length_ != input.buffer_length()
tom@419:       || channel_count_ != input.channel_count()) {
tom@419:     LOG_ERROR(_T("Mismatch between input to Initialize() and input to "
tom@419:                  "Process() in module %s."), module_identifier_.c_str());
tom@419:     return;
tom@419:   }
tom@419: 
tom@419:   int box_index = 0;
tom@419:   for (int c = 0; c < static_cast<int>(box_limits_channels_.size()); ++c) {
tom@419:     for (int s = 0; s < static_cast<int>(box_limits_time_.size()); ++s) {
tom@419:       int pixel_size_channels = (box_limits_channels_[c].first
tom@419:                                  - box_limits_channels_[c].second)
tom@419:                                    / box_size_spectral_;
tom@419:       int pixel_size_samples = box_limits_time_[s] / box_size_temporal_;
tom@419:       vector<vector<float> > box;
tom@419:       vector<float> line;
tom@419:       line.resize(box_size_temporal_, 0.0f);
tom@419:       box.resize(box_size_spectral_, line);
tom@419:       for (int i = 0; i < box_size_spectral_; ++i) {
tom@419:         for (int j = 0; j < box_size_temporal_; ++j) {
tom@419:           float pixel_value = 0.0f;
tom@419:           for (int k = i * pixel_size_channels;
tom@419:                k < (i + 1) * pixel_size_channels; ++k) {
tom@419:             for (int l = j * pixel_size_samples;
tom@419:                  l < (j + 1) * pixel_size_samples; ++l) {
tom@419:               pixel_value += input.sample(k
tom@419:                                           + box_limits_channels_[c].second, l);
tom@419:             }
tom@419:           }
tom@419:           pixel_value /= (pixel_size_channels * pixel_size_samples);
tom@419:           box[i][j] = pixel_value;
tom@419:         }
tom@419:       }
tom@419:       int feature_index = 0;
tom@419:       for (int i = 0; i < box_size_spectral_; ++i) {
tom@419:         float feature_value = 0.0f;
tom@419:         for (int j = 0; j < box_size_temporal_; ++j) {
tom@419:           feature_value += box[i][j];
tom@419:         }
tom@419:         feature_value /= box_size_temporal_;
tom@419:         output_.set_sample(box_index, feature_index, feature_value);
tom@419:         ++feature_index;
tom@419:       }
tom@419:       for (int j = 0; j < box_size_temporal_; ++j) {
tom@419:         float feature_value = 0.0f;
tom@419:         for (int i = 0; i < box_size_spectral_; ++i) {
tom@419:           feature_value += box[i][j];
tom@419:         }
tom@419:         feature_value /= box_size_spectral_;
tom@419:         output_.set_sample(box_index, feature_index, feature_value);
tom@419:         ++feature_index;
tom@419:       }
tom@419:       ++box_index;
tom@419:     }
tom@419:   }
tom@419: 
tom@419:   PushOutput();
tom@419: }
tom@419: }  // namespace aimc