tomwalters@277: // Copyright 2006-2010, Thomas Walters
tomwalters@277: //
tomwalters@277: // AIM-C: A C++ implementation of the Auditory Image Model
tomwalters@277: // http://www.acousticscale.org/AIMC
tomwalters@277: //
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@277: //
tomwalters@318: //     http://www.apache.org/licenses/LICENSE-2.0
tomwalters@277: //
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@277: 
tomwalters@277: /*!
tomwalters@277:  * \file
tomwalters@279:  * \brief File output in the HTK format.
tomwalters@277:  *
tomwalters@279:  * \author Tom Walters <tom@acousticscale.org>
tomwalters@277:  * \author Willem van Engen <cnbh@willem.engen.nl>
tomwalters@277:  * \date created 2006/10/30
tomwalters@277:  * \version \$Id$
tomwalters@277:  */
tomwalters@277: 
tomwalters@277: #ifdef _WINDOWS
tomwalters@320: #  include <direct.h>  // for _mkdir & _rmdir
tomwalters@277: #else
tomwalters@279: #  include <sys/types.h>
tomwalters@320: #  include <dirent.h>  // for opendir & friends
tomwalters@277: #endif
tomwalters@293: 
tomwalters@293: #include <stdint.h>
tomwalters@277: #include <stdio.h>
tomwalters@277: #include <string.h>
tomwalters@277: #include <cmath>
tomwalters@277: 
tomwalters@277: #include "Modules/Output/FileOutputHTK.h"
tomwalters@277: 
tomwalters@277: namespace aimc {
tomwalters@277: FileOutputHTK::FileOutputHTK(Parameters *params) : Module(params) {
tomwalters@278:   module_description_ = "File output in HTK format";
tomwalters@278:   module_identifier_ = "htk_out";
tomwalters@278:   module_type_ = "output";
tomwalters@278:   module_version_ = "$Id$";
tomwalters@402:   
tomwalters@402:   file_suffix_ = parameters_->DefaultString("htk_out.file_suffix", ".htk");
tomwalters@278: 
tomwalters@279:   file_handle_ = NULL;
tomwalters@279:   header_written_ = false;
tomwalters@277:   frame_period_ms_ = 0.0f;
tomwalters@402:   previous_start_time_ = 0;
tomwalters@277: }
tomwalters@277: 
tomwalters@277: FileOutputHTK::~FileOutputHTK() {
tomwalters@279:   if (file_handle_ != NULL)
tomwalters@279:     CloseFile();
tomwalters@277: }
tomwalters@277: 
tomwalters@402: bool FileOutputHTK::InitializeInternal(const SignalBank &input) {
tomwalters@402:   channel_count_ = input.channel_count();
tomwalters@402:   buffer_length_ = input.buffer_length();
tomwalters@402:   ResetInternal();
tomwalters@402:   if (file_handle_ == NULL) {
tomwalters@402:     LOG_ERROR(_T("Couldn't initialize file output."));
tomwalters@279:     return false;
tomwalters@279:   }
tomwalters@402:   if (!header_written_) {
tom@421:     WriteHeader(channel_count_ * buffer_length_);
tomwalters@292:   }
tomwalters@402:   
tomwalters@277:   return true;
tomwalters@277: }
tomwalters@277: 
tomwalters@277: void FileOutputHTK::ResetInternal() {
tomwalters@402:   // Finalize and close the open file, if there is one.
tomwalters@277:   if (file_handle_ != NULL && !header_written_) {
tom@421:     WriteHeader(channel_count_ * buffer_length_);
tomwalters@277:   }
tomwalters@292:   if (file_handle_ != NULL)
tomwalters@292:     CloseFile();
tomwalters@402:     
tomwalters@402:   // Now open and set up the new file.
tomwalters@402:   // Check that the output file exists and is writeable.
tomwalters@402:   string out_filename;
tomwalters@402:   out_filename = global_parameters_->GetString("output_filename_base") + file_suffix_;
tomwalters@402:   if ((file_handle_ = fopen(out_filename.c_str(),
tomwalters@402:                             "wb")) == NULL) {
tomwalters@402:     LOG_ERROR(_T("Couldn't open output file '%s' for writing."),
tomwalters@402:               out_filename.c_str());
tomwalters@402:     return;
tomwalters@402:   }
tomwalters@402:   sample_count_ = 0;
tomwalters@402:   header_written_ = false;
tom@421:   WriteHeader(channel_count_ * buffer_length_);
tomwalters@277: }
tomwalters@277: 
tom@421: void FileOutputHTK::WriteHeader(int num_elements) {
tomwalters@279:   if (header_written_)
tomwalters@279:     return;
tomwalters@277: 
tomwalters@279:   /* HTK format file: (taken from the HTK book - section 5.10.1)
tomwalters@279:    * Header: 12 bytes in total, contains:
tomwalters@279:    * sample_count - number of samples in file (4-byte integer)(long)
tomwalters@279:    * sample_period - sample period in 100ns units (4-byte integer)(long)
tomwalters@279:    * sample_size - number of bytes per sample (2-byte integer)(short)
tomwalters@279:    * parameter_kind - a code indicating the sample kind (2-byte integer)(short)
tomwalters@279:    */
tomwalters@277: 
tomwalters@279:   // To be filled in when the file is done
tomwalters@279:   int32_t sample_count = 0;
tomwalters@277: 
tom@421:   int32_t sample_period = floor(1e4 * frame_period_ms_);
tomwalters@289:   int16_t sample_size = num_elements * sizeof(float);  // NOLINT
tomwalters@277: 
tomwalters@277:   // User-defined coefficients with energy term
tomwalters@279:   int16_t parameter_kind = H_USER + H_E;
tomwalters@277: 
tomwalters@279:   // Fix endianness
tomwalters@279:   sample_count = ByteSwap32(sample_count);
tomwalters@279:   sample_period = ByteSwap32(sample_period);
tomwalters@279:   sample_size = ByteSwap16(sample_size);
tomwalters@279:   parameter_kind = ByteSwap16(parameter_kind);
tomwalters@277: 
tomwalters@277:   // Enter header values. sample_count is a dummy value which is filled in on
tomwalters@277:   // file close
tomwalters@279:   fwrite(&sample_count, sizeof(sample_count), 1, file_handle_);
tomwalters@279:   fwrite(&sample_period, sizeof(sample_period), 1, file_handle_);
tomwalters@279:   fwrite(&sample_size, sizeof(sample_size), 1, file_handle_);
tomwalters@279:   fwrite(&parameter_kind, sizeof(parameter_kind), 1, file_handle_);
tomwalters@279:   fflush(file_handle_);
tomwalters@277: 
tomwalters@279:   header_written_ = true;
tomwalters@277: }
tomwalters@277: 
tomwalters@277: void FileOutputHTK::Process(const SignalBank &input) {
tomwalters@277:   if (file_handle_ == NULL) {
tomwalters@277:     LOG_ERROR(_T("Couldn't process file output. No file is open."
tomwalters@277:                  "Please call FileOutputHTK::OpenFile first"));
tomwalters@279:     return;
tomwalters@277:   }
tomwalters@277: 
tomwalters@277:   if (!header_written_) {
tomwalters@292:     LOG_ERROR(_T("No header has been written on the output file yet. Please "
tomwalters@402:                  "call FileOutputHTK::Initialize() or FileOutputHTK::Reset() "
tom@442:                  "before calling FileOutputHTK::Process()"));
tomwalters@279:     return;
tomwalters@277:   }
tomwalters@279:   float s;
tomwalters@277: 
tomwalters@279:   for (int ch = 0; ch < input.channel_count(); ch++) {
tomwalters@277:     for (int i = 0; i < input.buffer_length(); i++) {
tomwalters@277:       s = input.sample(ch, i);
tomwalters@277:       s = ByteSwapFloat(s);
tomwalters@280:       fwrite(&s, sizeof(s), 1, file_handle_);
tomwalters@277:     }
tomwalters@277:   }
tomwalters@279:   sample_count_++;
tomwalters@402:   frame_period_ms_ = 1000.0
tomwalters@402:                      * (input.start_time() - previous_start_time_)
tomwalters@402:                      / input.sample_rate();
tomwalters@402:   previous_start_time_ = input.start_time();
tomwalters@277: }
tomwalters@277: 
tomwalters@277: bool FileOutputHTK::CloseFile() {
tomwalters@279:   if (file_handle_ == NULL)
tomwalters@277:     return false;
tomwalters@277: 
tomwalters@279:   // Write the first 4 bytes of the file
tomwalters@279:   // with how many samples there are in the file
tom@421:   // and the next 4 bytes with the frame period.
tomwalters@279:   fflush(file_handle_);
tomwalters@279:   rewind(file_handle_);
tomwalters@279:   fflush(file_handle_);
tomwalters@277:   int32_t samples = sample_count_;
tomwalters@279:   samples = ByteSwap32(samples);
tom@421:   int32_t sample_period = floor(1e4 * frame_period_ms_);
tom@421:   sample_period = ByteSwap32(sample_period);
tomwalters@279:   fwrite(&samples, sizeof(samples), 1, file_handle_);
tom@421:   fwrite(&sample_period, sizeof(sample_period), 1, file_handle_);
tomwalters@277: 
tomwalters@279:   // And close the file
tomwalters@279:   fclose(file_handle_);
tomwalters@279:   file_handle_ = NULL;
tomwalters@292:   header_written_ = false;
tomwalters@279:   return true;
tomwalters@277: }
tomwalters@277: 
tomwalters@277: float FileOutputHTK::ByteSwapFloat(float d) {
tomwalters@277:   // Endianness fix
tomwalters@277:   float a;
tomwalters@277:   unsigned char *dst = (unsigned char *)&a;
tomwalters@277:   unsigned char *src = (unsigned char *)&d;
tomwalters@277: 
tomwalters@277:   dst[0] = src[3];
tomwalters@277:   dst[1] = src[2];
tomwalters@277:   dst[2] = src[1];
tomwalters@277:   dst[3] = src[0];
tomwalters@277: 
tomwalters@277:   return a;
tomwalters@277: }
tomwalters@280: }  // namespace aimc
tomwalters@277: