tomwalters@280: // Copyright 2006-2010, Thomas Walters, Willem van Engen
tomwalters@280: //
tomwalters@280: // AIM-C: A C++ implementation of the Auditory Image Model
tomwalters@280: // http://www.acousticscale.org/AIMC
tomwalters@280: //
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@280: //
tomwalters@318: //     http://www.apache.org/licenses/LICENSE-2.0
tomwalters@280: //
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@280: 
tomwalters@277: /*!
tomwalters@277:  * \file
tomwalters@281:  * \brief File output to HTK format
tomwalters@277:  *
tomwalters@283:  * \author Thomas Walters <tom@acousticscale.org>
tomwalters@280:  * \author Willem van Engen <cnbh@willem.engen.nl>
tomwalters@277:  * \date created 2006/10/30
tomwalters@277:  * \version \$Header$
tomwalters@277:  */
tomwalters@280: 
tomwalters@283: #ifndef AIMC_MODULES_OUTPUT_HTK_H_
tomwalters@283: #define AIMC_MODULES_OUTPUT_HTK_H_
tomwalters@277: 
tomwalters@402: #include <string>
tomwalters@402: 
tomwalters@277: #include "Support/Module.h"
tomwalters@277: #include "Support/SignalBank.h"
tomwalters@277: 
tomwalters@402: // Defines from HTKwrite.c and The HTK Book
tomwalters@280: #define H_WAVEFORM    0  // sampled waveform
tomwalters@280: #define H_LPC         1  // linear prediction filter coefficients
tomwalters@280: #define H_LPREFC      2  // linear prediction reflection coefficients
tomwalters@280: #define H_LPCEPSTRA   3  // LPC cepstral coefficients
tomwalters@280: #define H_LPDELCEP    4  // LPC cepstra plus delta coefficients
tomwalters@280: #define H_IREFC       5  // LPC reflection coef in 16 bit integer format
tomwalters@280: #define H_MFCC        6  // mel-frequency cepstral coefficients
tomwalters@280: #define H_FBANK       7  // log mel-filter bank channel outputs
tomwalters@280: #define H_MELSPEC     8  // linear mel-filter bank channel outputs
tomwalters@280: #define H_USER        9  // user defined sample kind
tomwalters@280: #define H_DISCRETE    10  // vector quantised data
tomwalters@280: #define H_PLP         11  // Perceptual Linear Prediction
tomwalters@280: #define H_ANON        12  // Anonymous
tomwalters@277: 
tomwalters@280: #define H_E 64  // has energy
tomwalters@280: #define H_N 128  // absolute energy suppressed
tomwalters@280: #define H_D 256  // has delta coefficients
tomwalters@280: #define H_A 512  // has acceleration coefficients
tomwalters@280: #define H_C 1024  // is compressed
tomwalters@280: #define H_Z 2048  // has zero mean static coef.
tomwalters@280: #define H_K 4096  // has CRC checksum
tomwalters@280: #define H_O 8192  // has 0th cepstral coef.
tomwalters@280: #define H_V 16384  // Attach vq index
tomwalters@280: #define H_T 32768  // Attach delta-delta-delta index
tomwalters@277: 
tomwalters@277: // HTK fomat is big-endian...
tomwalters@277: #define ByteSwap16(n) \
tomwalters@277: ( ((((uint16_t) n) << 8) & 0xFF00) | \
tomwalters@277: ((((uint16_t) n) >> 8) & 0x00FF) )
tomwalters@277: 
tomwalters@277: #define ByteSwap32(n) \
tomwalters@277: ( ((((uint32_t) n) << 24) & 0xFF000000) | \
tomwalters@277: ((((uint32_t) n) << 8) & 0x00FF0000) | \
tomwalters@277: ((((uint32_t) n) >> 8) & 0x0000FF00) | \
tomwalters@277: ((((uint32_t) n) >> 24) & 0x000000FF) )
tomwalters@277: 
tomwalters@277: namespace aimc {
tomwalters@402: using std::string;
tomwalters@277: class FileOutputHTK : public Module {
tomwalters@277:  public:
tomwalters@279:   /*! \brief Create a new file output for an HTK format file. Use of this 
tomwalters@279:    *  class only really makes sense for the output of 1-D frames.
tomwalters@279:    */
tomwalters@280:   explicit FileOutputHTK(Parameters *pParam);
tomwalters@279:   ~FileOutputHTK();
tomwalters@277: 
tomwalters@279:   /*! \brief Initialize the output to HTK.
tomwalters@279:    *  \param *filename Filename of the ouptut file to be created.
tomwalters@279:    *  If the file exists it will be overwritten
tomwalters@279:    *  \return Returns true on success of initialization.
tomwalters@279:    */
tomwalters@277:   bool CloseFile();
tomwalters@277:   virtual void Process(const SignalBank &input);
tomwalters@279:  private:
tomwalters@277:   virtual bool InitializeInternal(const SignalBank &input);
tomwalters@277:   virtual void ResetInternal();
tomwalters@277: 
tomwalters@277:   float ByteSwapFloat(float d);
tomwalters@277: 
tom@421:   void WriteHeader(int nelements);
tomwalters@277: 
tomwalters@280:   /*! \brief Whether initialization is done or not
tomwalters@280:    */
tomwalters@279:   bool header_written_;
tomwalters@277: 
tomwalters@280:   /*! \brief Internal pointer to the output file
tomwalters@280:    */
tomwalters@279:   FILE *file_handle_;
tomwalters@277: 
tomwalters@280:   /*! \brief Count of the number of samples in the file, written on close
tomwalters@280:    */
tomwalters@279:   int sample_count_;
tomwalters@402:   string file_suffix_;
tomwalters@277: 
tomwalters@277:   int channel_count_;
tomwalters@277:   int buffer_length_;
tomwalters@277:   float frame_period_ms_;
tomwalters@402:   int previous_start_time_;
tomwalters@277: };
tomwalters@277: }  // namespace aimc
tomwalters@277: 
tomwalters@283: #endif  // AIMC_MODULES_OUTPUT_HTK_H_
tomwalters@277: