[FFmpeg-devel] [PATCH] avfilter: add arnndn filter

Lynne dev at lynne.ee
Sat Jul 27 16:54:48 EEST 2019


Jul 26, 2019, 4:37 PM by onemda at gmail.com <mailto:onemda at gmail.com>:
> +static void biquad(float *y, float mem[2], const float *x,
> +                   const float *b, const float *a, int N)
> +{
> +    for (int i = 0; i < N; i++) {
> +        float xi, yi;
> +        xi = x[i];
> +        yi = x[i] + mem[0];
> +        mem[0] = mem[1] + (b[0]*(double)xi - a[0]*(double)yi);
> +        mem[1] = (b[1]*(double)xi - a[1]*(double)yi);
> +        y[i] = yi;
> +    }
> +}

Do you really need double precision here?


> +static void apply_window(float *x, float *half_window)
> +{
> +    for (int i = 0; i < FRAME_SIZE; i++) {
> +        x[i] *= half_window[i];
> +        x[WINDOW_SIZE - 1 - i] *= half_window[i];
> +    }
> +}

Use vector_fmul_window.


> +#define FRAME_SIZE_SHIFT 2
> +#define FRAME_SIZE (120<<FRAME_SIZE_SHIFT)
> +#define WINDOW_SIZE (2*FRAME_SIZE)
> +#define FREQ_SIZE (FRAME_SIZE + 1)
> 
> +    AVComplexFloat y[WINDOW_SIZE];
>
> +    st->tx_fn(st->tx, y, x, 0);
> +
> +    for (int i = 0; i < FREQ_SIZE; i++) {
> +        out[i] = y[i];
> +    }

Make y and x aligned and set the tx stride to sizeof(float) for when SIMD is implemented.
I can't see how this could even work with stride = 0.
Why is FREQ_SIZE = FRAME_SIZE + 1?


> +    st->tx_fn(st->tx, y, x, 0);
> +
> +    /* output in reverse order for IFFT. */
> +    out[0] = y[0].re / WINDOW_SIZE;
> +    for (i=1; i < WINDOW_SIZE; i++) {
> +        out[i] = y[WINDOW_SIZE - i].re / WINDOW_SIZE;
> +    }

Make another tx context for an inverse transform, also same as above, and use float_dsp to divide all by WINDOW_SIZE.


> +    for (int i=0;i<FRAME_SIZE;i++)
> +        out[i] = x[i] + st->synthesis_mem[i] 

float_dsp? If you can afford to overwrite and overread the arrays odd sizes will work.


> +    for (int i = 0; i < FRAME_SIZE; i++)
> +        s->half_window[i] = sin(.5*M_PI*sin(.5*M_PI*(i+.5)/FRAME_SIZE) * sin(.5*M_PI*(i+.5)/FRAME_SIZE));

Just copy ff_celt_window and use it.


Also fix coding style and upload a weights file somewhere.


More information about the ffmpeg-devel mailing list