Mercurial > libavcodec.hg
changeset 11847:67206bbbab10 libavcodec
Rewrite the SBR decoder QMF analysis filter on top of the IMDCT instead of the RDFT.
This is based on a procedure to implement the filterbank on the DCT-IV described in
Hsu, H.W. et al. "A Complex Quadrature Mirror Filterbanks for MPEG-4 HE-AAC"
http://sites.google.com/site/wenchiehlee1020/Papers/ComplexFB_AES_121paper144.pdf
| author | alexc |
|---|---|
| date | Mon, 07 Jun 2010 21:52:08 +0000 |
| parents | 44a08df9971c |
| children | 6b52f0c3263d |
| files | aacsbr.c sbr.h |
| diffstat | 2 files changed, 17 insertions(+), 29 deletions(-) [+] |
line wrap: on
line diff
--- a/aacsbr.c Mon Jun 07 21:34:56 2010 +0000 +++ b/aacsbr.c Mon Jun 07 21:52:08 2010 +0000 @@ -71,9 +71,6 @@ static VLC vlc_sbr[10]; static const int8_t vlc_sbr_lav[10] = { 60, 60, 24, 24, 31, 31, 12, 12, 31, 12 }; -static DECLARE_ALIGNED(16, float, analysis_cos_pre)[64]; -static DECLARE_ALIGNED(16, float, analysis_sin_pre)[64]; -static DECLARE_ALIGNED(16, float, analysis_cossin_post)[32][2]; static const DECLARE_ALIGNED(16, float, zero64)[64]; #define SBR_INIT_VLC_STATIC(num, size) \ @@ -116,16 +113,6 @@ SBR_INIT_VLC_STATIC(8, 592); SBR_INIT_VLC_STATIC(9, 512); - for (n = 0; n < 64; n++) { - float pre = M_PI * n / 64; - analysis_cos_pre[n] = cosf(pre); - analysis_sin_pre[n] = sinf(pre); - } - for (k = 0; k < 32; k++) { - float post = M_PI * (k + 0.5) / 128; - analysis_cossin_post[k][0] = 4.0 * cosf(post); - analysis_cossin_post[k][1] = -4.0 * sinf(post); - } for (n = 1; n < 320; n++) sbr_qmf_window_us[320 + n] = sbr_qmf_window_us[320 - n]; sbr_qmf_window_us[384] = -sbr_qmf_window_us[384]; @@ -142,13 +129,13 @@ sbr->data[0].synthesis_filterbank_samples_offset = SBR_SYNTHESIS_BUF_SIZE - (1280 - 128); sbr->data[1].synthesis_filterbank_samples_offset = SBR_SYNTHESIS_BUF_SIZE - (1280 - 128); ff_mdct_init(&sbr->mdct, 7, 1, 1.0/64); - ff_rdft_init(&sbr->rdft, 6, IDFT_R2C); + ff_mdct_init(&sbr->mdct_ana, 7, 1, -2.0); } av_cold void ff_aac_sbr_ctx_close(SpectralBandReplication *sbr) { ff_mdct_end(&sbr->mdct); - ff_rdft_end(&sbr->rdft); + ff_mdct_end(&sbr->mdct_ana); } static int qsort_comparison_function_int16(const void *a, const void *b) @@ -1139,7 +1126,7 @@ * @param x pointer to the beginning of the first sample window * @param W array of complex-valued samples split into subbands */ -static void sbr_qmf_analysis(DSPContext *dsp, RDFTContext *rdft, const float *in, float *x, +static void sbr_qmf_analysis(DSPContext *dsp, RDFTContext *mdct, const float *in, float *x, float z[320], float W[2][32][32][2], float scale) { @@ -1156,19 +1143,20 @@ dsp->vector_fmul_reverse(z, sbr_qmf_window_ds, x, 320); for (k = 0; k < 64; k++) { float f = z[k] + z[k + 64] + z[k + 128] + z[k + 192] + z[k + 256]; - z[k] = f * analysis_cos_pre[k]; - z[k+64] = f; + z[k] = f; } - ff_rdft_calc(rdft, z); - re = z[0] * 0.5f; - im = 0.5f * dsp->scalarproduct_float(z+64, analysis_sin_pre, 64); - W[1][i][0][0] = re * analysis_cossin_post[0][0] - im * analysis_cossin_post[0][1]; - W[1][i][0][1] = re * analysis_cossin_post[0][1] + im * analysis_cossin_post[0][0]; + //Shuffle to IMDCT + z[64] = z[0]; for (k = 1; k < 32; k++) { - re = z[2*k ] - re; - im = z[2*k+1] - im; - W[1][i][k][0] = re * analysis_cossin_post[k][0] - im * analysis_cossin_post[k][1]; - W[1][i][k][1] = re * analysis_cossin_post[k][1] + im * analysis_cossin_post[k][0]; + z[64+2*k-1] = z[ k]; + z[64+2*k ] = -z[64-k]; + } + z[64+63] = z[32]; + + ff_imdct_half(mdct, z, z+64); + for (k = 0; k < 32; k++) { + W[1][i][k][0] = -z[63-k]; + W[1][i][k][1] = z[k]; } x += 32; } @@ -1730,7 +1718,7 @@ } for (ch = 0; ch < nch; ch++) { /* decode channel */ - sbr_qmf_analysis(&ac->dsp, &sbr->rdft, ch ? R : L, sbr->data[ch].analysis_filterbank_samples, + sbr_qmf_analysis(&ac->dsp, &sbr->mdct_ana, ch ? R : L, sbr->data[ch].analysis_filterbank_samples, (float*)sbr->qmf_filter_scratch, sbr->data[ch].W, 1/(-1024 * ac->sf_scale)); sbr_lf_gen(ac, sbr, sbr->X_low, sbr->data[ch].W);