diff --git a/libavcodec/aarch64/Makefile b/libavcodec/aarch64/Makefile
index ec1fd1eccc..21d169559c 100644
--- a/libavcodec/aarch64/Makefile
+++ b/libavcodec/aarch64/Makefile
@@ -3,6 +3,7 @@ OBJS-$(CONFIG_H264CHROMA)               += aarch64/h264chroma_init_aarch64.o
 OBJS-$(CONFIG_H264DSP)                  += aarch64/h264dsp_init_aarch64.o
 OBJS-$(CONFIG_H264QPEL)                 += aarch64/h264qpel_init_aarch64.o
 OBJS-$(CONFIG_HPELDSP)                  += aarch64/hpeldsp_init_aarch64.o
+OBJS-$(CONFIG_MPEGAUDIODSP)             += aarch64/mpegaudiodsp_init.o
 OBJS-$(CONFIG_NEON_CLOBBER_TEST)        += aarch64/neontest.o
 OBJS-$(CONFIG_VIDEODSP)                 += aarch64/videodsp_init.o
 
@@ -18,4 +19,5 @@ NEON-OBJS-$(CONFIG_H264DSP)             += aarch64/h264dsp_neon.o              \
 NEON-OBJS-$(CONFIG_H264QPEL)            += aarch64/h264qpel_neon.o             \
                                            aarch64/hpeldsp_neon.o
 NEON-OBJS-$(CONFIG_HPELDSP)             += aarch64/hpeldsp_neon.o
+NEON-OBJS-$(CONFIG_MPEGAUDIODSP)        += aarch64/mpegaudiodsp_neon.o
 NEON-OBJS-$(CONFIG_MDCT)                += aarch64/mdct_neon.o
diff --git a/libavcodec/aarch64/mpegaudiodsp_init.c b/libavcodec/aarch64/mpegaudiodsp_init.c
new file mode 100644
index 0000000000..b94514645f
--- /dev/null
+++ b/libavcodec/aarch64/mpegaudiodsp_init.c
@@ -0,0 +1,39 @@
+/*
+ * 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 <stdint.h>
+
+#include "libavutil/attributes.h"
+#include "libavutil/aarch64/cpu.h"
+#include "libavcodec/mpegaudiodsp.h"
+#include "config.h"
+
+void ff_mpadsp_apply_window_fixed_neon(int32_t *synth_buf, int32_t *window,
+                                       int *dither, int16_t *samples, int incr);
+void ff_mpadsp_apply_window_float_neon(float *synth_buf, float *window,
+                                       int *dither, float *samples, int incr);
+
+av_cold void ff_mpadsp_init_aarch64(MPADSPContext *s)
+{
+    int cpu_flags = av_get_cpu_flags();
+
+    if (have_neon(cpu_flags)) {
+        s->apply_window_fixed = ff_mpadsp_apply_window_fixed_neon;
+        s->apply_window_float = ff_mpadsp_apply_window_float_neon;
+    }
+}
diff --git a/libavcodec/aarch64/mpegaudiodsp_neon.S b/libavcodec/aarch64/mpegaudiodsp_neon.S
new file mode 100644
index 0000000000..e540eb8c01
--- /dev/null
+++ b/libavcodec/aarch64/mpegaudiodsp_neon.S
@@ -0,0 +1,226 @@
+/*
+ * Copyright (c) 2014 Janne Grunau <janne-libav@jannau.net>
+ *
+ * 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 "libavutil/aarch64/asm.S"
+
+#define FRAC_BITS   23   // fractional bits for sb_samples and dct
+#define WFRAC_BITS  16   // fractional bits for window
+#define OUT_SHIFT (WFRAC_BITS + FRAC_BITS - 15)
+
+const   tbl_rev128.s align=4
+        .byte           12, 13, 14, 15
+        .byte            8,  9, 10, 11
+        .byte            4,  5,  6,  7
+        .byte            0,  1,  2,  3
+endconst
+
+.macro   apply_window   type, st
+function ff_mpadsp_apply_window_\type\()_neon, export=1
+        mov             x7,  x0
+        sxtw            x4,  w4  // incr
+        add             x8,  x0,  #512<<2
+        ld1             {v0.4s,v1.4s,v2.4s,v3.4s},  [x7],  #64
+        ld1             {v4.4s,v5.4s,v6.4s,v7.4s},  [x7],  #64
+        st1             {v0.4s,v1.4s,v2.4s,v3.4s},  [x8],  #64
+        st1             {v4.4s,v5.4s,v6.4s,v7.4s},  [x8],  #64
+        movrel          x15, tbl_rev128.s
+        ld1             {v27.4s}, [x15]
+.ifc \type, fixed
+        lsl             x4,  x4,  #1
+.else
+        lsl             x4,  x4,  #2
+.endif
+        add             x10, x0,  #45<<2
+        add             x0,  x0,  #16<<2
+        add             x1,  x1,  #16<<2
+        add             x5,  x3,  x4,  lsl #5
+        sub             x5,  x5,  x4            // samples2
+        neg             x13, x4                 // -incr
+        mov             x9,  #64<<2
+.ifc \type, fixed
+        ld1r            {v16.2s}, [x2]          // dither_state
+        sxtl            v16.2d, v16.2s
+        movi            v29.2d, #0
+        movi            v30.2d, #(1<<OUT_SHIFT)-1
+        trn1            v31.2d, v29.2d, v30.2d
+        trn2            v30.2d, v30.2d, v29.2d
+        trn1            v16.2d, v16.2d, v29.2d
+.else
+        movi            v16.4s, #0
+        movi            v28.4s, #0
+.endif
+        mov             x14, #4
+1:
+        mov             x8,  x0
+        sub             x7,  x1,  #3<<2
+        sub             x6,  x1,  x14, lsl #4
+        add             x7,  x7,  x14, lsl #4
+        add             x11, x6, #(32)<<2      // w  + 32
+        add             x12, x7, #(32)<<2      // w2 + 32
+        mov             x15, #8
+        movi            v17.2d, #0
+        movi            v18.2d, #0
+        movi            v19.2d, #0
+2:
+        subs            x15, x15, #1
+        ld1             {v0.4s},  [x8],  x9
+        ld1             {v1.4s},  [x10], x9
+        ld1             {v2.4s},  [x6],  x9
+        ld1             {v3.4s},  [x7],  x9
+        tbl             v6.16b, {v0.16b}, v27.16b
+        tbl             v7.16b, {v1.16b}, v27.16b
+        ld1             {v4.4s},  [x11], x9
+        ld1             {v5.4s},  [x12], x9
+        MLA             v16, v2, v0
+        MLA2            v17, v2, v0
+        MLS             v18, v3, v6
+        MLS2            v19, v3, v6
+        MLS             v16, v4, v7
+        MLS2            v17, v4, v7
+        MLS             v18, v5, v1
+        MLS2            v19, v5, v1
+        b.gt            2b
+
+        cmp             x14, #4
+        sub             x10, x10, #64<<5        // 64 * 8 * sizeof(int32_t)
+
+.ifc \type, fixed
+        and             v28.16b, v16.16b, v30.16b
+        ext             v28.16b, v29.16b, v28.16b, #8
+
+        b.eq            4f
+        round_sample    v19, 1, 1
+4:
+        round_sample    v16, 1, 0
+        shrn            v16.2s, v16.2d,  #OUT_SHIFT
+        round_sample    v19, 0, 0
+        shrn            v19.2s, v19.2d,  #OUT_SHIFT
+        round_sample    v17, 0, 1
+        round_sample    v18, 1, 1
+        round_sample    v17, 1, 0
+        shrn2           v16.4s, v17.2d,  #OUT_SHIFT
+        round_sample    v18, 0, 0
+        shrn2           v19.4s, v18.2d,  #OUT_SHIFT
+        sqxtn           v16.4h, v16.4s
+        sqxtn           v18.4h, v19.4s
+.else
+        ext             v18.16b, v18.16b, v18.16b, #8
+.endif
+
+        st1             {v16.\st\()}[0], [x3], x4
+        b.eq            4f
+        st1             {v18.\st\()}[1], [x5], x13
+4:
+        st1             {v16.\st\()}[1], [x3], x4
+        st1             {v18.\st\()}[0], [x5], x13
+        st1             {v16.\st\()}[2], [x3], x4
+        st1             {v18.\st\()}[3], [x5], x13
+        st1             {v16.\st\()}[3], [x3], x4
+        st1             {v18.\st\()}[2], [x5], x13
+
+        mov             v16.16b, v28.16b
+
+        subs            x14, x14, #1
+        add             x0,  x0,  #4<<2
+        sub             x10, x10, #4<<2
+        b.gt            1b
+
+// comuting samples[16]
+        add             x6,  x1,  #32<<2
+        ld1             {v0.2s},  [x6],  x9
+        ld1             {v1.2s},  [x0],  x9
+.rept   3
+        ld1             {v2.2s},  [x6],  x9
+        ld1             {v3.2s},  [x0],  x9
+        MLS             v16, v0,  v1
+        ld1             {v0.2s},  [x6],  x9
+        ld1             {v1.2s},  [x0],  x9
+        MLS             v16, v2,  v3
+.endr
+        ld1             {v2.2s},  [x6],  x9
+        ld1             {v3.2s},  [x0],  x9
+        MLS             v16, v0,  v1
+        MLS             v16, v2,  v3
+
+.ifc \type, fixed
+        and             v28.16b, v16.16b, v30.16b
+        shrn            v20.2s,  v16.2d,  #OUT_SHIFT
+        xtn             v28.2s,  v28.2d
+        sqxtn           v20.4h,  v20.4s
+        st1             {v28.s}[0], [x2]        // save dither_state
+        st1             {v20.h}[0], [x3]
+.else
+        st1             {v16.s}[0], [x3]
+.endif
+
+        ret
+endfunc
+.purgem round_sample
+.purgem MLA
+.purgem MLA2
+.purgem MLS
+.purgem MLS2
+.endm
+
+
+.macro  round_sample    r, idx, next
+        add             \r\().2d, \r\().2d, v28.2d
+.if \idx == 0
+        and             v28.16b,  \r\().16b,  v30.16b
+.else // \idx == 1
+        and             v28.16b,  \r\().16b,  v31.16b
+.endif
+.if \idx != \next
+  .if \next == 0
+        ext             v28.16b, v28.16b, v29.16b, #8
+  .else
+        ext             v28.16b, v29.16b, v28.16b, #8
+  .endif
+.endif
+.endm
+.macro  MLA             d, s1, s2
+        smlal           \d\().2d, \s1\().2s, \s2\().2s
+.endm
+.macro  MLA2            d, s1, s2
+        smlal2          \d\().2d, \s1\().4s, \s2\().4s
+.endm
+.macro  MLS             d, s1, s2
+        smlsl           \d\().2d, \s1\().2s, \s2\().2s
+.endm
+.macro  MLS2            d, s1, s2
+        smlsl2          \d\().2d, \s1\().4s, \s2\().4s
+.endm
+apply_window fixed, h
+
+
+// nothing to do for round_sample and ML{A,S}2
+.macro  round_sample    r, idx, next
+.endm
+.macro  MLA2            d, s1, s2
+.endm
+.macro  MLS2            d, s1, s2
+.endm
+.macro  MLA             d, s1, s2
+        fmla            \d\().4s, \s1\().4s, \s2\().4s
+.endm
+.macro  MLS             d, s1, s2
+        fmls            \d\().4s, \s1\().4s, \s2\().4s
+.endm
+apply_window float, s
diff --git a/libavcodec/mpegaudiodsp.c b/libavcodec/mpegaudiodsp.c
index ddb7428458..5fe3444896 100644
--- a/libavcodec/mpegaudiodsp.c
+++ b/libavcodec/mpegaudiodsp.c
@@ -41,6 +41,7 @@ av_cold void ff_mpadsp_init(MPADSPContext *s)
     s->imdct36_blocks_float = ff_imdct36_blocks_float;
     s->imdct36_blocks_fixed = ff_imdct36_blocks_fixed;
 
+    if (ARCH_AARCH64) ff_mpadsp_init_aarch64(s);
     if (ARCH_ARM)     ff_mpadsp_init_arm(s);
     if (ARCH_PPC)     ff_mpadsp_init_ppc(s);
     if (ARCH_X86)     ff_mpadsp_init_x86(s);
diff --git a/libavcodec/mpegaudiodsp.h b/libavcodec/mpegaudiodsp.h
index 38c75c7fe5..a722a2f36b 100644
--- a/libavcodec/mpegaudiodsp.h
+++ b/libavcodec/mpegaudiodsp.h
@@ -55,6 +55,7 @@ void ff_mpa_synth_filter_float(MPADSPContext *s,
                                float *samples, int incr,
                                float *sb_samples);
 
+void ff_mpadsp_init_aarch64(MPADSPContext *s);
 void ff_mpadsp_init_arm(MPADSPContext *s);
 void ff_mpadsp_init_ppc(MPADSPContext *s);
 void ff_mpadsp_init_x86(MPADSPContext *s);