findsts.java

FreeTTS is a speech synthesis system written entirely in the Java programming language. It is based

     * Reconstruct a wave from a wave, sts, and lpc
     *
     * @param sampleRate the sample rate to use
     * @param lpc lpc
     * @param lpc_min minimum lpc value
     * @param lpc_range range of lpc values
     */
    public Wave(int sampleRate, STS[] stsData, LPC lpc, float lpc_min,
            float lpc_range) {
        // set number of samples and sample rate
        numSamples = 0;
        for (int i = 0; i < lpc.getNumFrames(); i++) {
            numSamples += stsData[i].getNumSamples();
        }
        samples = new short[numSamples];
        this.sampleRate = sampleRate;

        int start = 0;
        int end;
        int[] lpcResTimes = new int[lpc.getNumFrames()];
        int[] lpcResSizes = new int[lpc.getNumFrames()];
        short[] lpcResResidual = new short[numSamples];
        int[][] lpcResFrames = new int[lpc.getNumFrames()][];
        int lpcResNumChannels = lpc.getNumChannels() - 1;

        // load initial data
        for (int i = 0; i < lpc.getNumFrames(); i++) {
            lpcResTimes[i] = (int) (lpc.getTime(i) * sampleRate);
            lpcResFrames[i] = stsData[i].getFrame();
            end = start + stsData[i].getNumSamples();
            lpcResSizes[i] = stsData[i].getNumSamples();
            start = end;
        }

        for (int r = 0, i = 0; i < lpc.getNumFrames(); i++) {
            for (int j = 0; j < stsData[i].getNumSamples(); j++, r++) {
                lpcResResidual[r] = stsData[i].getResidual(j);
            }
        }

        float[] lpcCoefs = new float[lpcResNumChannels];
        float[] outbuf = new float[lpcResNumChannels + 1];
        int ci, cr;
        //float pp = 0;  // the C code uses this unnecessarily (for now)

        for (int r = 0, o = lpcResNumChannels, i = 0; i <
                lpc.getNumFrames(); i++) {
            // residual_fold is hard-coded to 1.
            int pm_size_samps = lpcResSizes[i];//  * residual_fold;

            // Unpack the LPC coefficients
            for (int k = 0; k < lpcResNumChannels; k++) {
                lpcCoefs[k] = (float)
                    ((((double) lpcResFrames[i][k])/65535.0) * lpc_range)
                    + lpc_min;
            }


            // resynthesize the signal
            for (int j = 0; j < pm_size_samps; j++, r++) {
                outbuf[o] = (float)
                    Utility.ulawToShort(lpcResResidual[r/* /residual_fold */]);

                cr = (o == 0 ? lpcResNumChannels : o-1);
                for (ci = 0; ci < lpcResNumChannels; ci++) {
                        outbuf[o] += lpcCoefs[ci] * outbuf[cr];
                        cr = (cr == 0 ? lpcResNumChannels : cr - 1);
                }
                samples[r] = (short) (outbuf[o]
                    /* + pp * lpcres->post_emphasis)*/); // post_emphasis = 0
                // pp = outbuf[o];
                o = (o == lpcResNumChannels ? 0 : o+1);
            }
        }
    }

    /**
     * Compare two waves and output how close the two are.
     * Useful for checking the general accuracy of find sts.
     *
     * <p>
     * Output may not exactly match that of flite find_sts
     * on Intel platforms due to discrepencies in the way that
     * Intel Pentiums perform floating point computations.
     * </p>
     *
     * @param the wave to compare this wave against
     *
     */
    public void compare(Wave wave2) {
        if (numSamples > wave2.numSamples) {
            wave2.compare(this);
        } else {
            double r = 0;
            int i = 0;
            for (i = 0; i < this.numSamples; i++) {
                r += (double)((float)this.samples[i] - (float)wave2.samples[i])
                    *(double)((float)this.samples[i] - (float)wave2.samples[i]);
            }
            r /= this.numSamples;
            System.out.println("a/b diff " + Double.toString(StrictMath.sqrt(r)));
        }
    }

    /**
     * Make sure that a string of characters appear next in the file
     *
     * @param dis DataInputStream to read in
     * @param chars a String containing the ascii characters you
     *          want the <code>dis</code> to contain.
     *
     * @return <code>true</code> if <code>chars</code> appears next
     *          in <code>dis</code>, else <code>false</code>
     * @throws on ill-formatted input (end of file, for example)
     */
    private boolean checkChars(DataInputStream dis, String chars)
            throws IOException {
        char[] carray = chars.toCharArray();
        for (int i = 0; i < carray.length; i++) {
            if ((char) dis.readByte() != carray[i]) {
                return false;
            }
        }
        return true;
    }

    /**
     * Get the sample rate for this wave
     *
     * @return sample rate
     */
    public int getSampleRate() {
        return sampleRate;
    }

    /**
     * Get the number of samples for this wave
     *
     * @return number of samples
     */
    public int getNumSamples() {
        return numSamples;
    }

    /* Get the sample data of this wave
     *
     * @return samples
     */
    public short[] getSamples() {
        return samples;
    }
}

/**
 * The sts data
 */
class STS {
    private int[] frame;
    private int numSamples;
    private short[] residual;

    /**
     * Create an empty STS
     */
    public STS() {
    }

    /**
     * Create an sts with the given data
     *
     * @param frame frame for this sts
     * @param numSamples number of samples this sts will contain
     * @param residual the residual for this sts
     * 
     */
    public STS(int[] frame, int numSamples, short[] residual) {
        this.frame = new int[frame.length];
        System.arraycopy(frame, 0, this.frame, 0, frame.length);
        this.numSamples = numSamples;
        this.residual = new short[residual.length];
        System.arraycopy(residual, 0, this.residual, 0, residual.length);
    }

    /**
     * Get the number of samples associated with this sts
     *
     * @return the number of samples for this sts
     */
    public int getNumSamples() {
        return numSamples;
    }

    /**
     * Get the residual associated with this sts
     *
     * @return residual associated with this sts
     */
    public short getResidual(int i) {
        return residual[i];
    }

    /**
     * Get the frame associated with this sts
     *
     * @return a copy of the frame associated with this sts
     */
    public int[] getFrame() {
        int[] f = new int[frame.length];
        System.arraycopy(frame, 0, f, 0, frame.length);
        return f;
    }

    /**
     * Get an entry out of the frame
     *
     * @param index the index into the frame
     *
     * @return the entry in the frame at offset <code>index</code>
     */
    public int getFrameEntry(int index) {
        return frame[index];
    }
}


/**
 * This class is for general purpose functions such as reading and
 * writing from files, or converting formats of numbers.
 */
class Utility {

    /**
     * Reads the next word (text separated by whitespace) from the
     * given stream
     *
     * @param dis the input stream
     *
     * @return the next word
     *
     * @throws IOException on error
     */
    public static String readWord(DataInputStream dis) throws IOException {
        StringBuffer sb = new StringBuffer();
        char c;

        // skip leading whitespace
        do {
            c = readChar(dis);
        } while(Character.isWhitespace(c));

        // read the word
        do {
            sb.append(c);
            c = readChar(dis);
        } while (!Character.isWhitespace(c));
        return sb.toString();
    }

    /**
     * Reads a single char from the stream
     *
     * @param dis the stream to read
     * @return the next character on the stream
     *
     * @throws IOException if an error occurs
     */
    public static char readChar(DataInputStream dis) throws IOException {
        return (char) dis.readByte();
    }

    /**
     * Reads a given number of chars from the stream
     *
     * @param dis the stream to read
     * @param num the number of chars to read
     * @return a character array containing the next <code>num<code>
     *          in the stream
     *
     * @throws IOException if an error occurs
     */
    public static char[] readChars(DataInputStream dis, int num)
            throws IOException {
        char[] carray = new char[num];
        for (int i = 0; i < num; i++) {
            carray[i] = readChar(dis);
        }
        return carray;
    }

    /**
     * Read a float from the input stream, byte-swapping as
     * necessary
     *
     * @param dis the inputstream
     * @param isBigEndian whether or not the data being read in is in
     *          big endian format.
     *
     * @return a floating pint value
     *
     * @throws IOException on error
     */
    public static float readFloat(DataInputStream dis, boolean isBigEndian)
            throws IOException {
        float val;
        if (!isBigEndian) {
            val =  readLittleEndianFloat(dis);
        } else {
            val =  dis.readFloat();
        }
        return val;
    }

    /**
     * Reads the next float from the given DataInputStream,
     * where the data is in little endian.
     *
     * @param dataStream the DataInputStream to read from
     *
     * @return a float
     */
    public static float readLittleEndianFloat(DataInputStream dataStream)
            throws IOException {
        return Float.intBitsToFloat(readLittleEndianInt(dataStream));
    }

    /**
     * Read an integer from the input stream, byte-swapping as
     * necessary
     *
     * @param dis the inputstream
     * @param isBigEndian whether or not the data being read in is in
     *          big endian format.
     *
     * @return an integer value
     *
     * @throws IOException on error
     */
    public static int readInt(DataInputStream dis, boolean isBigEndian)
            throws IOException {
        if (!isBigEndian) {
            return readLittleEndianInt(dis);
        } else {
            return dis.readInt();
        }
    }

    /**
     * Reads the next little-endian integer from the given DataInputStream.
     *
     * @param dataStream the DataInputStream to read from
     *
     * @return an integer
     */
    public static int readLittleEndianInt(DataInputStream dataStream)
            throws IOException {
        int bits = 0x00000000;
        for (int shift = 0; shift < 32; shift += 8) {
            int byteRead = (0x000000ff & dataStream.readByte());
            bits |= (byteRead << shift);
        }
        return bits;
    }

    /**
     * Read a short from the input stream, byte-swapping as
     * necessary
     *
     * @param dis the inputstream
     * @param isBigEndian whether or not the data being read in is in
     *          big endian format.
     *
     * @return an integer value
     *
     * @throws IOException on error
     */
    public static short readShort(DataInputStream dis, boolean isBigEndian)
        throws IOException {
        if (!isBigEndian) {
            return readLittleEndianShort(dis);
        } else {
            return dis.readShort();
        }
    }

    /**
     * Reads the next little-endian short from the given DataInputStream.
     *
     * @param dataStream the DataInputStream to read from
     *
     * @return a short
     */
    public static short readLittleEndianShort(DataInputStream dis)
        throws IOException {
        short bits = (short)(0x0000ff & dis.readByte());
        bits |= (((short)(0x0000ff & dis.readByte())) << 8);
        return bits;
    }

    /**
     * Convert a short to ulaw format
     * 
     * @param sample the short to convert
     *
     * @return a short containing an unsigned 8-bit quantity
     *          representing the ulaw
     */
    public static short shortToUlaw(short sample) {
        final int[] exp_lut = {0,0,1,1,2,2,2,2,3,3,3,3,3,3,3,3,
                                   4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,
                                   5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,
                                   5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,
                                   6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,
                                   6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,
                                   6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,
                                   6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,
                                   7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,
                                   7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,
                                   7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,
                                   7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,
                                   7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,
                                   7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,
                                   7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,
                                   7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7};

        int sign, exponent, mantissa;
        short ulawbyte;

        final short CLIP = 32635;
        final short BIAS = 0x0084;

        /* Get the sample into sign-magnitude. */
        sign = (sample >> 8) & 0x80; /* set aside the sign */
        if ( sign != 0 ) {
            sample = (short) -sample; /* get magnitude */
        }
        if ( sample > CLIP ) sample = CLIP; /* clip the magnitude */

        /* Convert from 16 bit linear to ulaw. */
        sample = (short) (sample + BIAS);
        exponent = exp_lut[( sample >> 7 ) & 0xFF];
        mantissa = ( sample >> ( exponent + 3 ) ) & 0x0F;
        ulawbyte = (short)
            ((~ ( sign | ( exponent << 4 ) | mantissa)) & 0x00FF);
        if ( ulawbyte == 0 ) ulawbyte = 0x02; /* optional CCITT trap */
        return ulawbyte;
    }

    /**
     * Convert a ulaw format to short
     * 
     * @param ulawbyte a short containing an unsigned 8-but quantity
     *          representing a ulaw
     *
     * @return the short equivalent of the ulaw
     */
    public static short ulawToShort(short ulawbyte) {
        final int[] exp_lut = { 0, 132, 396, 924, 1980, 4092, 8316, 16764 };
        int sign, exponent, mantissa;
        short sample;

        ulawbyte = (short) (ulawbyte & 0x00FF);
        ulawbyte = (short) (~ulawbyte);
        sign = ( ulawbyte & ((short) 0x80) );
        exponent = (int) ( (ulawbyte & (short) 0x00FF) >> 4 ) & 0x07;
        mantissa = ulawbyte & (short) 0x0F;
        sample = (short) (exp_lut[exponent] + (mantissa << (exponent + 3)));
        if ( sign != 0 ) sample = (short) (-sample);

        return sample;
    }


    /**
     * Print a float type's internal bit representation in hex
     *
     * @param f the float to print
     *
     * @return a string containing the hex value of <code>f</code>
     */
    public static String hex(float f) {
        return Integer.toHexString(Float.floatToIntBits(f));
    }
}