findsts.java

使用Exlipse编写的一个语音程序

/**
 * Portions Copyright 2003 Sun Microsystems, Inc.
 * Portions Copyright 1999-2003 Language Technologies Institute,
 * Carnegie Mellon University.
 * All Rights Reserved.  Use is subject to license terms.
 *
 * See the file "license.terms" for information on usage and
 * redistribution of this file, and for a DISCLAIMER OF ALL
 * WARRANTIES.
 *
 */

import java.io.FileInputStream;
import java.io.DataInputStream;
import java.io.OutputStreamWriter;
import java.io.FileOutputStream;
import java.io.IOException;
import java.io.FileNotFoundException;

import javax.sound.sampled.spi.AudioFileReader;
import javax.sound.sampled.AudioInputStream;


/**
 * Performs the generation of STS files in FestVox to FreeTTS
 * conversion.
 *
 * <p>
 * This program is a port from flite/tools/find_sts_main.c
 * </p>
 *
 * <p>
 * Note: 
 * The a/b diff result is slightly different than the C version due to
 * Intel floating-point math.
 * </p>
 */
public strictfp class FindSTS {
    /**
     * Generate an sts file
     *
     * <p>
     * <b>Usage</b>
     * </p>
     * <p>
     *  <code> java FindSTS lpc_min lpc_range lpcFile waveFile stsFile
     * </code>
     * </p>
     * <p>
     * <code> stsFile </code> the output file.
     * </p>
     */
    public static void main(String[] args) {
        try {
            if (args.length != 5) {
                System.err.println("Usage: java FindSTS lpc_min lpc_range "
                        + "lpcFile waveFile stsFile");
            } else {
                float lpc_min = Float.parseFloat(args[0]);
                float lpc_range = Float.parseFloat(args[1]);
                FileInputStream lpcFile = new FileInputStream(args[2]);
                FileInputStream waveFile = new FileInputStream(args[3]);
                FileOutputStream stsFile = new FileOutputStream(args[4]);

                // Read input
                LPC lpc = new LPC(new DataInputStream(lpcFile));
                Wave wave = new Wave(new DataInputStream(waveFile));
                lpcFile.close();
                waveFile.close();

                // Generate sts data
                STS[] stsData = findSTS(wave, lpc, lpc_min, lpc_range);

                // Verify STS data for sanity
                Wave reconstructedWave = new
                    Wave(wave.getSampleRate(), stsData, lpc,
                            lpc_min, lpc_range);
                wave.compare(reconstructedWave);

                // Save output
                OutputStreamWriter stsWriter = new OutputStreamWriter(stsFile);
                saveSTS(stsData, lpc, wave, stsWriter, lpc_min, lpc_range);

                stsWriter.close();
                stsFile.close();
            }
        } catch (FileNotFoundException ioe) {
            throw new Error("Error while running FindSTS" + ioe.getMessage());
        } catch (IOException ioe) {
            throw new Error("IO error while finding sts" + ioe.getMessage());
        }
    }

    /**
     * Find the sts data.
     *
     * @param wave the data from the wave file
     * @param lpc the data from the lpc file
     * @param lpc_min the minimum lpc value
     * @param lpc_range the range of the lpc values
     *
     * @return an <code>STS</code> array containing the data
     */
    private static STS[] findSTS(Wave wave, LPC lpc, float lpc_min,
            float lpc_range) {
        int size;
        int start = 0;
        int end;

        STS[] stsData = new STS[lpc.getNumFrames()];

        // read wave data into a special array.
        short[] waveData =
            new short[wave.getNumSamples() + lpc.getNumChannels()];
        System.arraycopy(wave.getSamples(), 0, waveData,
                    lpc.getNumChannels(), wave.getNumSamples());

        for (int i = 0; i < lpc.getNumFrames(); i++) {
            double[] resd;
            int[] frame;
            short[] residual;

            end = (int) ((float) wave.getSampleRate() * lpc.getTime(i));
            size = end - start;
            if (size <= 0) {
                System.out.println("frame size at "
                        + Float.toString(lpc.getTime(i)) + " is "
                        + Integer.toString(size) + ".");
            }

            residual = generateResiduals(waveData,
                    start + lpc.getNumChannels(), lpc.getFrame(i),
                    lpc.getNumChannels(), size);

            frame = new int[lpc.getNumChannels() - 1];
            for (int j = 1; j < lpc.getNumChannels(); j++) {
                frame[j - 1] = (int)
                    ((((lpc.getFrameEntry(i, j) - lpc_min) / lpc_range))
                     * (float) 65535.0);
            }

            stsData[i] = new STS(frame, size, residual);
            start = end;
        }

        return stsData;
    }

    /**
     * Generate the residuals for this sts
     *
     * @param wave specially formatted wave data
     * @param start offset into the wave data
     * @param frame frame data from the lpc
     * @param order typically the number of lpc channels
     * @param size size of the residual
     *
     * @return sts residuals
     */
    private static short[] generateResiduals(short[] wave, int start,
            float[] frame, int order, int size) {
        double r;
        short[] residual = new short[size];
        for (int i = 0; i < order; i++) {
            r = wave[start + i];
            for (int j = 1; j < order; j++) {
                r -= frame[j] * ((double) wave[start + (i - j)]);
            }
            residual[i] = Utility.shortToUlaw((short) r);
        }
        for (int i = order; i < size; i++) {
            r = wave[start + i];
            for (int j = 1; j < order; j++) {
                r -= frame[j] * ((double) wave[start + (i - j)]);
            } 
            residual[i] = Utility.shortToUlaw((short) r);
        }
        return residual;
    }

    /**
     * Save the sts data
     *
     * @param stsData generated sts data
     * @param lpc data loaded from the lpc file
     * @param wave data loaded from the wave file
     * @param osw the OutputStreamWriter to write the sts data to
     * @param lpc_min minimum lpc value
     * @param lpc_range range of lpc values
     *
     */
    private static void saveSTS(STS[] stsData, LPC lpc, Wave wave,
            OutputStreamWriter osw, float lpc_min, float lpc_range) {
        try {
            osw.write("( " + Integer.toString(lpc.getNumFrames())
                    + " " + Integer.toString(lpc.getNumChannels() - 1)
                    + " " + Integer.toString(wave.getSampleRate())
                    + " " + Float.toString(lpc_min)
                    + " " + Float.toString(lpc_range)
                    + ")\n");
            for (int m=0, i=0; i < lpc.getNumFrames(); i++) {
                osw.write("( " + Float.toString(lpc.getTime(i)) + " (");

                // Use the following line instead to compare against
                // the flite find_sts output
                //osw.write("( " + Utility.hex(lpc.getTime(i)) + " (");

                for (int j = 1; j < lpc.getNumChannels(); j++) {
                    osw.write(" " +
                            Integer.toString(stsData[i].getFrameEntry(j - 1)));
                }
                osw.write(" ) "
                        + Integer.toString(stsData[i].getNumSamples())
                        + " ( ");
                for (int j = 0; j < stsData[i].getNumSamples(); j++) {
                        osw.write(" " +
                                Integer.toString(stsData[i].getResidual(j)));
                }
                osw.write(" ))\n");
            }
        } catch (IOException ioe) {
            throw new Error("IO error while writing sts." + ioe.getMessage());
        }
    }
}


/**
 * The lpc data
 *
 */
class LPC {
    private int numFrames;
    private int numChannels;
    float[] times;
    float[][] frames;

    /** Create lpc data from an input stream
     *
     * @param dis DataInputStream to read the lpc in from
     *
     */
    public LPC(DataInputStream dis) {
        try {
            if (!Utility.readWord(dis).equals("EST_File") ||
                    !Utility.readWord(dis).equals("Track")) {
                throw new Error("Lpc file not EST Track file");
            }

            boolean isBinary = false;
            boolean isBigEndian = false;

            // Read Header
            String token = Utility.readWord(dis);
            while (!token.equals("EST_Header_End")) {
                if (token.equals("DataType")) {
                    if (Utility.readWord(dis).equals("binary")) {
                        isBinary = true;
                    } else {
                        isBinary = false;
                    }
                } else if (token.equals("ByteOrder")) {
                    if (Utility.readWord(dis).equals("10")) {
                        isBigEndian = true;
                    } else {
                        isBigEndian = false;
                    }
                } else if (token.equals("NumFrames")) {
                    numFrames = Integer.parseInt(Utility.readWord(dis));
                } else if (token.equals("NumChannels")) {
                    numChannels = Integer.parseInt(Utility.readWord(dis));
                }
                // Ignore all other content in header

                token = Utility.readWord(dis);
            }

            times = new float[numFrames];
            frames = new float[numFrames][numChannels];

            // read data section
            if (isBinary) {
                loadBinaryData(dis, isBigEndian);
            } else {
                loadTextData(dis);
            }
        }
        catch (IOException ioe) {
            throw new Error("IO error while parsing lpc" + ioe.getMessage());
        }
    }

    /**
     * load the data section of the lpc file as ascii text
     *
     * @param dis DataInputStream to read from
     *
     * @throws IOException on ill-formatted input
     */
    private void loadTextData(DataInputStream dis) throws IOException {
        for (int f=0; f < numFrames; f++) {
            times[f] = Float.parseFloat(Utility.readWord(dis));
            Utility.readWord(dis);  // can be only 1
            for (int c=0; c < numChannels; c++) {
                    frames[f][c] = Float.parseFloat(Utility.readWord(dis));
            }
        }
    }

    /**
     * load the data section of the lpc file as ascii text
     *
     * @param dis DataInputStream to read from
     * @param isBigEndian whether or not the data in the file is in
     *          big endian byte order
     *
     * @throws IOException on ill-formatted input
     */
    private void loadBinaryData(DataInputStream dis, boolean isBigEndian)
            throws IOException {
        for (int f=0; f < numFrames; f++) {
            times[f] = Utility.readFloat(dis, isBigEndian);

            // Ignore the 'breaks' field
            Utility.readFloat(dis, isBigEndian);

            for (int c=0; c < numChannels; c++) {
                frames[f][c] = Utility.readFloat(dis, isBigEndian);
            }
        }
    }

    /**
     * Get the number of frames in this lpc
     *
     * @return number of frames in this lpc
     */
    public int getNumFrames() {
        return numFrames;
    }

    /**
     * Get the number of channels in this lpc
     *
     * @return number of channels in this lpc
     */
    public int getNumChannels() {
        return numChannels;
    }

    /**
     * Get the times associated with this lpc
     *
     * @return an array of times associated with this lpc
     */
    public float[] getTimes() {
        return times;
    }

    /**
     * Get an individual time associated with this lpc
     *
     * @param index index of time to get
     *
     * @return time value at given index
     */
    public float getTime(int index) {
        return times[index];
    }

    /**
     * Get an individual frame
     *
     * @param i index of frame
     *
     * @return the frame
     */
    public float[] getFrame(int i) {
        return frames[i];
    }

    /**
     * Get an individual frame entry
     *
     * @param i index of frame
     * @param j index into frame
     *
     * @return the frame entry in frame <code>i</code> at index
     *          <code>j</code>
     */
    public float getFrameEntry(int i, int j) {
        return frames[i][j];
    }
}


/**
 * The wave (riff) data
 */
class Wave {
    private int numSamples;
    private int sampleRate;
    private short[] samples;

    // Only really used in loading of data.
    private int headerSize;
    private int numBytes;
    private int numChannels = 1;  // Only support mono

    static final short RIFF_FORMAT_PCM = 0x0001;

    /**
     * Read in a wave from a riff format
     *
     * @param dis DataInputStream to read data from
     */
    public Wave (DataInputStream dis) {
        try {
            loadHeader(dis);
            if (dis.skipBytes(headerSize - 16) != (headerSize - 16)) {
                throw new Error("Unexpected error parsing wave file.");
            }

            // Bunch of potential random headers
            while (true) {
                String s = new String(Utility.readChars(dis, 4));

                if (s.equals("data")) {
                    numSamples = Utility.readInt(dis, false) / 2;
                    break;
                } else if (s.equals("fact")) {
                    int i = Utility.readInt(dis, false);
                    if (dis.skipBytes(i) != i) {
                        throw new Error("Unexpected error parsing wave file.");
                    }
                } else {
                    throw new Error("Unsupported wave header chunk type " + s);
                }
            }

            int dataLength = numSamples * numChannels;
            samples = new short[numSamples];

            for (int i = 0; i < dataLength; i++) {
                samples[i] = Utility.readShort(dis, false);
            }

        } catch (IOException ioe) {
            throw new Error("IO error while parsing wave" + ioe.getMessage());
        }
    }

    /**
     * load a RIFF header
     *
     * @param dis DataInputStream to read from
     *
     * @throws IOException on ill-formatted input
     */
    private void loadHeader(DataInputStream dis) throws IOException {
        if (!checkChars(dis, "RIFF")) {
            throw new Error("Invalid wave file format.");
        }
        numBytes = Utility.readInt(dis,false);
        if (!checkChars(dis, "WAVEfmt ")) {
            throw new Error("Invalid wave file format.");
        }

        headerSize = Utility.readInt(dis, false);

        if (Utility.readShort(dis, false) != RIFF_FORMAT_PCM) {
            throw new Error("Invalid wave file format.");
        }

        if (Utility.readShort(dis, false) != 1) {
            throw new Error("Only mono wave files supported.");
        }
        
        sampleRate = Utility.readInt(dis, false);
        Utility.readInt(dis, false);
        Utility.readShort(dis, false);
        Utility.readShort(dis, false);
    }

    /**