various filters for ACELP-based codecs

Originally committed as revision 13110 to svn://svn.ffmpeg.org/ffmpeg/trunk

2 files changed, 238 insertions, 0 deletions

diff --git a/libavcodec/acelp_filters.c b/libavcodec/acelp_filters.c
new file mode 100644
index 0000000000..b98c6f741e
--- /dev/null
+++ b/libavcodec/acelp_filters.c
@@ -0,0 +1,125 @@
+/*
+ * various filters for ACELP-based codecs
+ *
+ * Copyright (c) 2008 Vladimir Voroshilov
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include <inttypes.h>
+
+#include "avcodec.h"
+#include "acelp_filters.h"
+#define FRAC_BITS 13
+#include "mathops.h"
+
+void ff_acelp_convolve_circ(
+        int16_t* fc_out,
+        const int16_t* fc_in,
+        const int16_t* filter,
+        int subframe_size)
+{
+    int i, k;
+
+    memset(fc_out, 0, subframe_size * sizeof(int16_t));
+
+    /* Since there are few pulses over entire subframe (i.e. almost all
+       fc_in[i] are zero, in case of G.729D the buffer contains two non-zero
+       samples before the call to ff_acelp_enhance_harmonics, and (due to
+       pitch_delay bounded to [20; 143]) a maximum four non-zero samples
+       for a total of 40 after the call to it), it is faster to swap two loops
+       and process non-zero samples only. This will reduce the number of
+       multiplications from 40*40 to 4*40 for G.729D */
+    for(i=0; i<subframe_size; i++)
+    {
+        if(fc_in[i])
+        {
+            for(k=0; k<i; k++)
+                fc_out[k] += (fc_in[i] * filter[subframe_size + k - i]) >> 15;
+
+            for(k=i; k<subframe_size; k++)
+                fc_out[k] += (fc_in[i] * filter[k - i]) >> 15;
+        }
+    }
+}
+
+int ff_acelp_lp_synthesis_filter(
+        int16_t *out,
+        const int16_t* filter_coeffs,
+        const int16_t* in,
+        int buffer_length,
+        int filter_length,
+        int stop_on_overflow)
+{
+    int i,n;
+
+    for(n=0; n<buffer_length; n++)
+    {
+        int sum = 0x800;
+        for(i=1; i<filter_length; i++)
+            sum -= filter_coeffs[i] * out[n-i];
+
+        sum = (sum >> 12) + in[n];
+
+        /* Check for overflow */
+        if(sum + 0x8000 > 0xFFFFU)
+        {
+            if(stop_on_overflow)
+                return 1;
+            sum = (sum >> 31) ^ 32767;
+        }
+        out[n] = sum;
+    }
+
+    return 0;
+}
+
+void ff_acelp_weighted_filter(
+        int16_t *out,
+        const int16_t* in,
+        const int16_t *weight_pow,
+        int filter_length)
+{
+    int n;
+    for(n=0; n<filter_length; n++)
+        out[n] = (in[n] * weight_pow[n] + 0x4000) >> 15; /* (3.12) = (0.15) * (3.12) with rounding */
+}
+
+void ff_acelp_high_pass_filter(
+        int16_t* out,
+        int hpf_f[2],
+        const int16_t* in,
+        int length)
+{
+    int i;
+    int tmp;
+
+    for(i=0; i<length; i++)
+    {
+        tmp =  MULL(hpf_f[0], 15836);                     /* (14.13) = (13.13) * (1.13) */
+        tmp += MULL(hpf_f[1], -7667);                     /* (13.13) = (13.13) * (0.13) */
+        tmp += 7699 * (in[i] - 2*in[i-1] + in[i-2]); /* (14.13) =  (0.13) * (14.0) */
+
+        /* Multiplication by 2 with rounding can cause short type
+           overflow, thus clipping is required. */
+
+        out[i] = av_clip_int16((tmp + 0x800) >> 12);      /* (15.0) = 2 * (13.13) = (14.13) */
+
+        hpf_f[1] = hpf_f[0];
+        hpf_f[0] = tmp;
+    }
+}
diff --git a/libavcodec/acelp_filters.h b/libavcodec/acelp_filters.h
new file mode 100644
index 0000000000..155a847a77
--- /dev/null
+++ b/libavcodec/acelp_filters.h
@@ -0,0 +1,113 @@
+/*
+ * various filters for ACELP-based codecs
+ *
+ * Copyright (c) 2008 Vladimir Voroshilov
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#ifndef FFMPEG_ACELP_FILTERS_H
+#define FFMPEG_ACELP_FILTERS_H
+
+/**
+ * \brief Circularly convolve fixed vector with a phase dispersion impulse
+ *        response filter (D.6.2 of G.729 and 6.1.5 of AMR).
+ * \param fc_out vector with filter applied
+ * \param fc_in source vector
+ * \param filter phase filter coefficients
+ *
+ *  fc_out[n] = sum(i,0,len-1){ fc_in[i] * filter[(len + n - i)%len] }
+ *
+ * \note fc_in and fc_out should not overlap!
+ */
+void ff_acelp_convolve_circ(
+        int16_t* fc_out,
+        const int16_t* fc_in,
+        const int16_t* filter,
+        int subframe_size);
+
+/**
+ * \brief LP synthesis filter
+ * \param out [out] pointer to output buffer
+ * \param filter_coeffs filter coefficients (-0x8000 <= (3.12) < 0x8000)
+ * \param in input signal
+ * \param buffer_length amount of data to process
+ * \param filter_length filter length (11 for 10th order LP filter)
+ * \param stop_on_overflow   1 - return immediately if overflow occurs
+ *                           0 - ignore overflows
+ *
+ * \return 1 if overflow occurred, 0 - otherwise
+ *
+ * \note Output buffer must contain 10 samples of past
+ *       speech data before pointer.
+ *
+ * Routine applies 1/A(z) filter to given speech data.
+ */
+int ff_acelp_lp_synthesis_filter(
+        int16_t *out,
+        const int16_t* filter_coeffs,
+        const int16_t* in,
+        int buffer_length,
+        int filter_length,
+        int stop_on_overflow);
+
+/**
+ * \brief Calculates coefficients of weighted A(z/weight) filter.
+ * \param out [out] weighted A(z/weight) result
+ *                  filter (-0x8000 <= (3.12) < 0x8000)
+ * \param in source filter (-0x8000 <= (3.12) < 0x8000)
+ * \param weight_pow array containing weight^i (-0x8000 <= (0.15) < 0x8000)
+ * \param filter_length filter length (11 for 10th order LP filter)
+ *
+ * out[i]=weight_pow[i]*in[i] , i=0..9
+ */
+void ff_acelp_weighted_filter(
+        int16_t *out,
+        const int16_t* in,
+        const int16_t *weight_pow,
+        int filter_length);
+
+/**
+ * \brief high-pass filtering and upscaling (4.2.5 of G.729)
+ * \param out [out] output buffer for filtered speech data
+ * \param hpf_f [in/out] past filtered data from previous (2 items long)
+ *                       frames (-0x20000000 <= (14.13) < 0x20000000)
+ * \param in speech data to process
+ * \param length input data size
+ *
+ * out[i] = 0.93980581 * in[i] - 1.8795834 * in[i-1] + 0.93980581 * in[i-2] +
+ *          1.9330735 * out[i-1] - 0.93589199 * out[i-2]
+ *
+ * The filter has a cut-off frequency of 100Hz
+ *
+ * \note Two items before the top of the out buffer must contain two items from the
+ *       tail of the previous subframe.
+ *
+ * \remark It is safe to pass the same array in in and out parameters
+ *
+ * \remark AMR uses mostly the same filter (cut-off frequency 60Hz, same formula,
+ *         but constants differs in 5th sign after comma). Fortunately in
+ *         fixed-point all coefficients are the same as in G.729. Thus this
+ *         routine can be used for the fixed-point AMR decoder, too.
+ */
+void ff_acelp_high_pass_filter(
+        int16_t* out,
+        int hpf_f[2],
+        const int16_t* in,
+        int length);
+
+#endif // FFMPEG_ACELP_FILTERS_H