ADPCM IMA WAV 2, 3 and 5 bits decoder

Signed-off-by: Paul B Mahol <onemda@gmail.com>

libavcodec/adpcm.c

@@ -118,10 +118,8 @@ static av_cold int adpcm_decode_init(AVCodecContext * avctx)
         c->status[0].step = c->status[1].step = 511;
         break;
     case AV_CODEC_ID_ADPCM_IMA_WAV:
-        if (avctx->bits_per_coded_sample != 4) {
+        if (avctx->bits_per_coded_sample < 2 || avctx->bits_per_coded_sample > 5)
-            av_log(avctx, AV_LOG_ERROR, "Only 4-bit ADPCM IMA WAV files are supported\n");
+            return AVERROR_INVALIDDATA;
-            return -1;
-        }
         break;
     case AV_CODEC_ID_ADPCM_IMA_APC:
         if (avctx->extradata && avctx->extradata_size >= 8) {
@@ -188,6 +186,29 @@ static inline short adpcm_ima_expand_nibble(ADPCMChannelStatus *c, char nibble,
     return (short)c->predictor;
 }
 
+static inline int16_t adpcm_ima_wav_expand_nibble(ADPCMChannelStatus *c, GetBitContext *gb, int bps)
+{
+    int nibble, step_index, predictor, sign, delta, diff, step, shift;
+
+    shift = bps - 1;
+    nibble = get_bits_le(gb, bps),
+    step = ff_adpcm_step_table[c->step_index];
+    step_index = c->step_index + ff_adpcm_index_tables[bps - 2][nibble];
+    step_index = av_clip(step_index, 0, 88);
+
+    sign = nibble & (1 << shift);
+    delta = nibble & ((1 << shift) - 1);
+    diff = ((2 * delta + 1) * step) >> shift;
+    predictor = c->predictor;
+    if (sign) predictor -= diff;
+    else predictor += diff;
+
+    c->predictor = av_clip_int16(predictor);
+    c->step_index = step_index;
+
+    return (int16_t)c->predictor;
+}
+
 static inline int adpcm_ima_qt_expand_nibble(ADPCMChannelStatus *c, int nibble, int shift)
 {
     int step_index;
@@ -557,10 +578,14 @@ static int get_nb_samples(AVCodecContext *avctx, GetByteContext *gb,
         nb_samples = (buf_size - 4 * ch) * 2 / ch;
         break;
     case AV_CODEC_ID_ADPCM_IMA_WAV:
+    {
+        int bsize = ff_adpcm_ima_block_sizes[avctx->bits_per_coded_sample - 2];
+        int bsamples = ff_adpcm_ima_block_samples[avctx->bits_per_coded_sample - 2];
         if (avctx->block_align > 0)
             buf_size = FFMIN(buf_size, avctx->block_align);
-        nb_samples = 1 + (buf_size - 4 * ch) / (4 * ch) * 8;
+        nb_samples = 1 + (buf_size - 4 * ch) / (bsize * ch) * bsamples;
         break;
+    }
     case AV_CODEC_ID_ADPCM_MS:
         if (avctx->block_align > 0)
             buf_size = FFMIN(buf_size, avctx->block_align);
@@ -720,6 +745,23 @@ static int adpcm_decode_frame(AVCodecContext *avctx, void *data,
             }
         }
 
+        if (avctx->bits_per_coded_sample != 4) {
+            int samples_per_block = ff_adpcm_ima_block_samples[avctx->bits_per_coded_sample - 2];
+            GetBitContext g;
+
+            init_get_bits8(&g, gb.buffer, bytestream2_get_bytes_left(&gb));
+            for (n = 0; n < (nb_samples - 1) / samples_per_block; n++) {
+                for (i = 0; i < avctx->channels; i++) {
+                    cs = &c->status[i];
+                    samples = &samples_p[i][1 + n * samples_per_block];
+                    for (m = 0; m < samples_per_block; m++) {
+                        samples[m] = adpcm_ima_wav_expand_nibble(cs, &g,
+                                          avctx->bits_per_coded_sample);
+                    }
+                }
+            }
+            bytestream2_skip(&gb, avctx->block_align - avctx->channels * 4);
+        } else {
         for (n = 0; n < (nb_samples - 1) / 8; n++) {
             for (i = 0; i < avctx->channels; i++) {
                 cs = &c->status[i];
@@ -731,6 +773,7 @@ static int adpcm_decode_frame(AVCodecContext *avctx, void *data,
                 }
             }
         }
+        }
         break;
     case AV_CODEC_ID_ADPCM_4XM:
         for (i = 0; i < avctx->channels; i++)

libavcodec/adpcm_data.c

@@ -27,12 +27,33 @@
 
 /* ff_adpcm_step_table[] and ff_adpcm_index_table[] are from the ADPCM
    reference source */
-/* This is the index table: */
+static const int8_t adpcm_index_table2[4] = {
+    -1,  2,
+    -1,  2,
+};
+
+static const int8_t adpcm_index_table3[8] = {
+    -1, -1,  1,  2,
+    -1, -1,  1,  2,
+};
+
 const int8_t ff_adpcm_index_table[16] = {
     -1, -1, -1, -1, 2, 4, 6, 8,
     -1, -1, -1, -1, 2, 4, 6, 8,
 };
 
+static const int8_t adpcm_index_table5[32] = {
+    -1, -1, -1, -1, -1, -1, -1, -1, 1, 2, 4, 6, 8, 10, 13, 16,
+    -1, -1, -1, -1, -1, -1, -1, -1, 1, 2, 4, 6, 8, 10, 13, 16,
+};
+
+const int8_t const *ff_adpcm_index_tables[4] = {
+    &adpcm_index_table2[0],
+    &adpcm_index_table3[0],
+    &ff_adpcm_index_table[0],
+    &adpcm_index_table5[0],
+};
+
 /**
  * This is the step table. Note that many programs use slight deviations from
  * this table, but such deviations are negligible:

libavcodec/adpcm_data.h

@@ -28,6 +28,10 @@
 
 #include <stdint.h>
 
+static const uint8_t ff_adpcm_ima_block_sizes[4]   = {  4, 12, 4, 20 };
+static const uint8_t ff_adpcm_ima_block_samples[4] = { 16, 32, 8, 32 };
+
+extern const int8_t const *ff_adpcm_index_tables[4];
 extern const int8_t  ff_adpcm_index_table[16];
 extern const int16_t ff_adpcm_step_table[89];
 extern const int16_t ff_adpcm_oki_step_table[49];

libavcodec/utils.c

@@ -2899,7 +2899,9 @@ int av_get_audio_frame_duration(AVCodecContext *avctx, int frame_bytes)
                 int blocks = frame_bytes / ba;
                 switch (avctx->codec_id) {
                 case AV_CODEC_ID_ADPCM_IMA_WAV:
-                    return blocks * (1 + (ba - 4 * ch) / (4 * ch) * 8);
+                    if (bps < 2 || bps > 5)
+                        return 0;
+                    return blocks * (1 + (ba - 4 * ch) / (bps * ch) * 8);
                 case AV_CODEC_ID_ADPCM_IMA_DK3:
                     return blocks * (((ba - 16) * 2 / 3 * 4) / ch);
                 case AV_CODEC_ID_ADPCM_IMA_DK4: