diff --git a/libavcodec/g729dec.c b/libavcodec/g729dec.c
index a0a2dba66be282c2faa7be60e9fe5b554af823b6..e8656bca5c560787ff26488985e051291f2dc83c 100644
--- a/libavcodec/g729dec.c
+++ b/libavcodec/g729dec.c
@@ -79,14 +79,6 @@ static inline int g729_get_parity(uint8_t value)
    return (0x6996966996696996ULL >> (value >> 2)) & 1;
 }
 
-        /*
-          This filter enhances harmonic components of the fixed-codebook vector to
-          improve the quality of the reconstructed speech.
-
-                     / fc_v[i],                                    i < pitch_delay
-          fc_v[i] = <
-                     \ fc_v[i] + gain_pitch * fc_v[i-pitch_delay], i >= pitch_delay
-        */
         ff_acelp_weighted_vector_sum(
                 fc + pitch_delay_int[i],
                 fc + pitch_delay_int[i],
@@ -102,16 +94,6 @@ static inline int g729_get_parity(uint8_t value)
             gain_corr_factor = cb_gain_1st_8k[parm->gc_1st_index[i]][1] +
                                cb_gain_2nd_8k[parm->gc_2nd_index[i]][1];
 
-        /* Routine requires rounding to lowest. */
-        ff_acelp_interpolate(
-                ctx->exc + i*ctx->subframe_size,
-                ctx->exc + i*ctx->subframe_size - pitch_delay_3x/3,
-                ff_acelp_interp_filter,
-                6,
-                (pitch_delay_3x%3)<<1,
-                10,
-                ctx->subframe_size);
-
         ff_acelp_weighted_vector_sum(
                 ctx->exc + i * ctx->subframe_size,
                 ctx->exc + i * ctx->subframe_size,