tomwalters@314: // Copyright 2008-2010, Thomas Walters tomwalters@314: // tomwalters@314: // AIM-C: A C++ implementation of the Auditory Image Model tomwalters@314: // http://www.acousticscale.org/AIMC tomwalters@314: // tomwalters@318: // Licensed under the Apache License, Version 2.0 (the "License"); tomwalters@318: // you may not use this file except in compliance with the License. tomwalters@318: // You may obtain a copy of the License at tomwalters@314: // tomwalters@318: // http://www.apache.org/licenses/LICENSE-2.0 tomwalters@314: // tomwalters@318: // Unless required by applicable law or agreed to in writing, software tomwalters@318: // distributed under the License is distributed on an "AS IS" BASIS, tomwalters@318: // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. tomwalters@318: // See the License for the specific language governing permissions and tomwalters@318: // limitations under the License. tomwalters@314: tomwalters@314: /*! tomwalters@314: * \file AIMCopy.cpp tomwalters@314: * \brief AIM-C replacement for HTK's HCopy tomwalters@314: * tomwalters@314: * The following subset of the command-line flags tomwalters@314: * should be implemented from HCopy: tomwalters@314: * -A Print command line arguments off tomwalters@314: * -C cf Set config file to cf default tomwalters@314: * (should be able to take multiple config files) tomwalters@314: * -S f Set script file to f none tomwalters@314: * //! \todo -T N Set trace flags to N 0 tomwalters@314: * -V Print version information off tomwalters@314: * -D of Write configuration data to of none tomwalters@314: * tomwalters@314: * \author Thomas Walters tomwalters@314: * \date created 2008/05/08 tomwalters@314: * \version \$Id$ tomwalters@314: */ tomwalters@314: tomwalters@314: #include tomwalters@314: #include tomwalters@314: #include tomwalters@314: #include tomwalters@314: #include tomwalters@314: tomwalters@314: #include tomwalters@314: #include tomwalters@314: tomwalters@314: #include "Modules/Input/ModuleFileInput.h" tomwalters@314: #include "Modules/BMM/ModuleGammatone.h" tomwalters@314: #include "Modules/BMM/ModulePZFC.h" tomwalters@314: #include "Modules/NAP/ModuleHCL.h" tomwalters@314: #include "Modules/Strobes/ModuleParabola.h" tomwalters@314: #include "Modules/Strobes/ModuleLocalMax.h" tomwalters@314: #include "Modules/SAI/ModuleSAI.h" tomwalters@314: #include "Modules/SSI/ModuleSSI.h" tomwalters@314: #include "Modules/SNR/ModuleNoise.h" tomwalters@314: #include "Modules/Profile/ModuleSlice.h" tomwalters@314: #include "Modules/Profile/ModuleScaler.h" tomwalters@314: #include "Modules/Features/ModuleGaussians.h" tomwalters@314: #include "Modules/Output/FileOutputHTK.h" tomwalters@314: #include "Support/Common.h" tomwalters@314: #include "Support/FileList.h" tomwalters@314: #include "Support/Parameters.h" tomwalters@314: tomwalters@314: using std::ofstream; tomwalters@314: using std::pair; tomwalters@314: using std::vector; tomwalters@314: using std::string; tomwalters@314: int main(int argc, char* argv[]) { tomwalters@314: string sound_file; tomwalters@314: string data_file; tomwalters@314: string config_file; tomwalters@314: string script_file; tomwalters@314: bool write_data = false; tomwalters@314: bool print_version = false; tomwalters@359: bool load_params = false; tomwalters@314: tomwalters@314: string version_string( tomwalters@314: " AIM-C AIMCopy\n" tomwalters@314: " (c) 2006-2010, Thomas Walters and Willem van Engen\n" tomwalters@314: " http://www.acoustiscale.org/AIMC/\n" tomwalters@314: "\n"); tomwalters@314: tomwalters@314: if (argc < 2) { tomwalters@314: printf("%s", version_string.c_str()); tomwalters@314: printf("AIMCopy is intended as a drop-in replacement for HTK's HCopy\n"); tomwalters@314: printf("command. It is used for making features from audio files for\n"); tomwalters@314: printf("use with HTK.\n"); tomwalters@314: printf("Usage: \n"); tomwalters@314: printf(" -A Print command line arguments off\n"); tomwalters@314: printf(" -C cf Set config file to cf none\n"); tomwalters@314: printf(" -S f Set script file to f none\n"); tomwalters@314: printf(" -V Print version information off\n"); tomwalters@314: printf(" -D g Write configuration data to g none\n"); tomwalters@314: return -1; tomwalters@314: } tomwalters@314: tomwalters@314: // Parse command-line arguments tomwalters@314: for (int i = 1; i < argc; i++) { tomwalters@314: if (strcmp(argv[i],"-A") == 0) { tomwalters@314: for (int j = 0; j < argc; j++) tomwalters@314: printf("%s ",argv[j]); tomwalters@314: printf("\n"); tomwalters@314: fflush(stdout); tomwalters@314: continue; tomwalters@314: } tomwalters@314: if (strcmp(argv[i],"-C") == 0) { tomwalters@314: if (++i >= argc) { tomwalters@314: aimc::LOG_ERROR(_T("Configuration file name expected after -C")); tomwalters@314: return(-1); tomwalters@314: } tomwalters@314: config_file = argv[i]; tomwalters@359: load_params = true; tomwalters@314: continue; tomwalters@314: } tomwalters@314: if (strcmp(argv[i],"-S") == 0) { tomwalters@314: if (++i >= argc) { tomwalters@314: aimc::LOG_ERROR(_T("Script file name expected after -S")); tomwalters@314: return(-1); tomwalters@314: } tomwalters@314: script_file = argv[i]; tomwalters@314: continue; tomwalters@314: } tomwalters@314: if (strcmp(argv[i],"-D") == 0) { tomwalters@314: if (++i >= argc) { tomwalters@314: aimc::LOG_ERROR(_T("Data file name expected after -D")); tomwalters@314: return(-1); tomwalters@314: } tomwalters@314: data_file = argv[i]; tomwalters@314: write_data = true; tomwalters@314: continue; tomwalters@314: } tomwalters@314: if (strcmp(argv[i],"-V") == 0) { tomwalters@314: print_version = true; tomwalters@314: continue; tomwalters@314: } tomwalters@314: aimc::LOG_ERROR(_T("Unrecognized command-line argument: %s"), argv[i]); tomwalters@314: } tomwalters@314: tomwalters@314: if (print_version) tomwalters@314: printf("%s", version_string.c_str()); tomwalters@314: tomwalters@314: aimc::Parameters params; tomwalters@314: tomwalters@359: if (load_params) { tomwalters@359: if (!params.Load(config_file.c_str())) { tomwalters@359: aimc::LOG_ERROR(_T("Couldn't load parameters from file %s"), tomwalters@359: config_file.c_str()); tomwalters@359: return -1; tomwalters@359: } tomwalters@314: } tomwalters@314: tomwalters@314: vector > file_list = aimc::FileList::Load(script_file); tomwalters@314: if (file_list.size() == 0) { tomwalters@314: aimc::LOG_ERROR("No data read from file %s", script_file.c_str()); tomwalters@314: return -1; tomwalters@314: } tomwalters@314: tomwalters@314: // Set up AIM-C processor here tomwalters@314: aimc::ModuleFileInput input(¶ms); tomwalters@314: //aimc::ModuleNoise noise_maker(¶ms); tomwalters@314: aimc::ModuleGammatone bmm(¶ms); tomwalters@314: aimc::ModuleHCL nap(¶ms); tomwalters@314: aimc::ModuleLocalMax strobes(¶ms); tomwalters@314: aimc::ModuleSAI sai(¶ms); tomwalters@314: params.SetBool("ssi.pitch_cutoff", false); tomwalters@314: aimc::ModuleSSI ssi_no_cutoff(¶ms); tomwalters@314: tomwalters@314: params.SetBool("ssi.pitch_cutoff", true); tomwalters@314: params.SetFloat("ssi.pitch_search_start_ms", 4.6f); tomwalters@314: aimc::ModuleSSI ssi_cutoff(¶ms); tomwalters@314: tomwalters@314: params.SetBool("slice.all", false); tomwalters@314: params.SetInt("slice.lower_index", 77); tomwalters@314: params.SetInt("slice.upper_index", 150); tomwalters@314: aimc::ModuleSlice slice_ssi_slice_1_no_cutoff(¶ms); tomwalters@314: aimc::ModuleSlice slice_ssi_slice_1_cutoff(¶ms); tomwalters@314: tomwalters@314: params.SetBool("slice.all", true); tomwalters@314: aimc::ModuleSlice slice_ssi_all_no_cutoff(¶ms); tomwalters@314: aimc::ModuleSlice slice_ssi_all_cutoff(¶ms); tomwalters@314: tomwalters@314: params.SetFloat("nap.lowpass_cutoff", 100.0); tomwalters@314: aimc::ModuleHCL smooth_nap(¶ms); tomwalters@314: params.SetBool("slice.all", true); tomwalters@314: aimc::ModuleSlice nap_profile(¶ms); tomwalters@314: aimc::ModuleScaler nap_scaler(¶ms); tomwalters@314: tomwalters@314: aimc::ModuleGaussians nap_features(¶ms); tomwalters@314: aimc::ModuleGaussians features_ssi_slice1_no_cutoff(¶ms); tomwalters@314: aimc::ModuleGaussians features_ssi_slice1_cutoff(¶ms); tomwalters@314: aimc::ModuleGaussians features_ssi_all_no_cutoff(¶ms); tomwalters@314: aimc::ModuleGaussians features_ssi_all_cutoff(¶ms); tomwalters@314: tomwalters@314: aimc::FileOutputHTK nap_out(¶ms); tomwalters@314: aimc::FileOutputHTK output_ssi_slice1_no_cutoff(¶ms); tomwalters@314: aimc::FileOutputHTK output_ssi_slice1_cutoff(¶ms); tomwalters@314: aimc::FileOutputHTK output_ssi_all_no_cutoff(¶ms); tomwalters@314: aimc::FileOutputHTK output_ssi_all_cutoff(¶ms); tomwalters@314: tomwalters@314: input.AddTarget(&bmm); tomwalters@314: //noise_maker.AddTarget(&bmm); tomwalters@314: bmm.AddTarget(&nap); tomwalters@314: bmm.AddTarget(&smooth_nap); tomwalters@314: smooth_nap.AddTarget(&nap_profile); tomwalters@314: nap_profile.AddTarget(&nap_scaler); tomwalters@314: nap_scaler.AddTarget(&nap_features); tomwalters@314: nap_features.AddTarget(&nap_out); tomwalters@314: tomwalters@314: nap.AddTarget(&strobes); tomwalters@314: strobes.AddTarget(&sai); tomwalters@314: sai.AddTarget(&ssi_no_cutoff); tomwalters@314: sai.AddTarget(&ssi_cutoff); tomwalters@314: tomwalters@315: ssi_no_cutoff.AddTarget(&slice_ssi_slice_1_no_cutoff); tomwalters@315: ssi_no_cutoff.AddTarget(&slice_ssi_all_no_cutoff); tomwalters@315: ssi_cutoff.AddTarget(&slice_ssi_slice_1_cutoff); tomwalters@315: ssi_cutoff.AddTarget(&slice_ssi_all_cutoff); tomwalters@314: tomwalters@314: slice_ssi_slice_1_no_cutoff.AddTarget(&features_ssi_slice1_no_cutoff); tomwalters@314: slice_ssi_all_no_cutoff.AddTarget(&features_ssi_all_no_cutoff); tomwalters@314: slice_ssi_slice_1_cutoff.AddTarget(&features_ssi_slice1_cutoff); tomwalters@314: slice_ssi_all_cutoff.AddTarget(&features_ssi_all_cutoff); tomwalters@314: tomwalters@314: tomwalters@314: features_ssi_slice1_no_cutoff.AddTarget(&output_ssi_slice1_no_cutoff); tomwalters@314: features_ssi_all_no_cutoff.AddTarget(&output_ssi_all_no_cutoff); tomwalters@314: features_ssi_slice1_cutoff.AddTarget(&output_ssi_slice1_cutoff); tomwalters@314: features_ssi_all_cutoff.AddTarget(&output_ssi_all_cutoff); tomwalters@314: tomwalters@314: tomwalters@314: if (write_data) { tomwalters@314: ofstream outfile(data_file.c_str()); tomwalters@314: if (outfile.fail()) { tomwalters@314: aimc::LOG_ERROR("Couldn't open data file %s for writing", tomwalters@314: data_file.c_str()); tomwalters@314: return -1; tomwalters@314: } tomwalters@314: time_t rawtime; tomwalters@314: struct tm * timeinfo; tomwalters@314: time(&rawtime); tomwalters@314: timeinfo = localtime(&rawtime); tomwalters@314: tomwalters@314: tomwalters@314: outfile << "# AIM-C AIMCopy\n"; tomwalters@314: outfile << "# Run on: " << asctime(timeinfo); tomwalters@314: char * descr = getenv("USER"); tomwalters@314: if (descr) { tomwalters@314: outfile << "# By user: " << descr <<"\n"; tomwalters@314: } tomwalters@323: outfile << "# Module chain:\n"; tomwalters@323: outfile << "#"; tomwalters@323: input.PrintTargets(outfile); tomwalters@323: outfile << "\n"; tomwalters@323: outfile << "#\n"; tomwalters@323: input.PrintVersions(outfile); tomwalters@323: outfile << "\n"; tomwalters@314: outfile << "#\n"; tomwalters@314: outfile << "# Parameters:\n"; tomwalters@314: outfile << params.WriteString(); tomwalters@314: outfile.close(); tomwalters@314: } tomwalters@314: tomwalters@314: for (unsigned int i = 0; i < file_list.size(); ++i) { tomwalters@314: // aimc::LOG_INFO(_T("In: %s"), file_list[i].first.c_str()); tomwalters@314: aimc::LOG_INFO(_T("Out: %s"), file_list[i].second.c_str()); tomwalters@314: tomwalters@359: string filename = file_list[i].second + "slice_1_no_cutoff"; tomwalters@314: output_ssi_slice1_no_cutoff.OpenFile(filename.c_str(), 10.0f); tomwalters@359: filename = file_list[i].second + "ssi_profile_no_cutoff"; tomwalters@314: output_ssi_all_no_cutoff.OpenFile(filename.c_str(), 10.0f); tomwalters@359: filename = file_list[i].second + "slice_1_cutoff"; tomwalters@314: output_ssi_slice1_cutoff.OpenFile(filename.c_str(), 10.0f); tomwalters@359: filename = file_list[i].second + "ssi_profile_cutoff"; tomwalters@314: output_ssi_all_cutoff.OpenFile(filename.c_str(), 10.0f); tomwalters@359: filename = file_list[i].second + "smooth_nap_profile"; tomwalters@314: nap_out.OpenFile(filename.c_str(), 10.0f); tomwalters@314: tomwalters@314: if (input.LoadFile(file_list[i].first.c_str())) { tomwalters@314: input.Process(); tomwalters@314: } else { tomwalters@314: printf("LoadFile failed for file %s\n", file_list[i].first.c_str()); tomwalters@314: } tomwalters@314: input.Reset(); tomwalters@314: } tomwalters@314: tomwalters@314: return 0; tomwalters@314: }