Adding a safeguard against unstable LPC, so now there's no way (that I can

think of) to produce NaNs in the new PLC.
This commit is contained in:
Jean-Marc Valin 2009-12-28 07:59:42 -05:00
parent d69c1cb37c
commit 07fed1bf76

View file

@ -1347,13 +1347,20 @@ static void celt_decode_lost(CELTDecoder * restrict st, celt_word16 * restrict p
iir(e, st->lpc, e, len+st->mode->overlap, LPC_ORDER, mem);
{
float ratio, S2=0;
float S2=0;
for (i=0;i<len+overlap;i++)
S2 += e[i]*1.*e[i];
ratio = sqrt((S1+1)/(S2+1));
if (ratio < 1)
/* This checks for an "explosion" in the synthesis (including NaNs) */
if (!(S1 > 0.2f*S2))
{
for (i=0;i<len+overlap;i++)
e[i] = 0;
} else if (S1 < S2)
{
float ratio = sqrt((S1+1)/(S2+1));
for (i=0;i<len+overlap;i++)
e[i] *= ratio;
}
}
for (i=0;i<MAX_PERIOD+st->mode->overlap-N;i++)