internal static float fast_atan2f(float y, float x)
{
float x2, y2;
/* Should avoid underflow on the values we'll get */
if (Inlines.ABS16(x) + Inlines.ABS16(y) < 1e-9f)
{
x *= 1e12f;
y *= 1e12f;
}
x2 = x * x;
y2 = y * y;
if (x2 < y2)
{
float den = (y2 + cB * x2) * (y2 + cC * x2);
if (den != 0)
{
return(-x * y * (y2 + cA * x2) / den + (y < 0 ? -cE : cE));
}
else
{
return(y < 0 ? -cE : cE);
}
}
else
{
float den = (x2 + cB * y2) * (x2 + cC * y2);
if (den != 0)
{
return(x * y * (x2 + cA * y2) / den + (y < 0 ? -cE : cE) - (x * y < 0 ? -cE : cE));
}
else
{
return((y < 0 ? -cE : cE) - (x * y < 0 ? -cE : cE));
}
}
}
internal static int compute_stereo_width(short[] pcm, int pcm_ptr, int frame_size, int Fs, StereoWidthState mem)
{
int corr;
int ldiff;
int width;
int xx, xy, yy;
int sqrt_xx, sqrt_yy;
int qrrt_xx, qrrt_yy;
int frame_rate;
int i;
int short_alpha;
frame_rate = Fs / frame_size;
short_alpha = CeltConstants.Q15ONE - (25 * CeltConstants.Q15ONE / Inlines.IMAX(50, frame_rate));
xx = xy = yy = 0;
for (i = 0; i < frame_size - 3; i += 4)
{
int pxx = 0;
int pxy = 0;
int pyy = 0;
int x, y;
int p2i = pcm_ptr + (2 * i);
x = pcm[p2i];
y = pcm[p2i + 1];
pxx = Inlines.SHR32(Inlines.MULT16_16(x, x), 2);
pxy = Inlines.SHR32(Inlines.MULT16_16(x, y), 2);
pyy = Inlines.SHR32(Inlines.MULT16_16(y, y), 2);
x = pcm[p2i + 2];
y = pcm[p2i + 3];
pxx += Inlines.SHR32(Inlines.MULT16_16(x, x), 2);
pxy += Inlines.SHR32(Inlines.MULT16_16(x, y), 2);
pyy += Inlines.SHR32(Inlines.MULT16_16(y, y), 2);
x = pcm[p2i + 4];
y = pcm[p2i + 5];
pxx += Inlines.SHR32(Inlines.MULT16_16(x, x), 2);
pxy += Inlines.SHR32(Inlines.MULT16_16(x, y), 2);
pyy += Inlines.SHR32(Inlines.MULT16_16(y, y), 2);
x = pcm[p2i + 6];
y = pcm[p2i + 7];
pxx += Inlines.SHR32(Inlines.MULT16_16(x, x), 2);
pxy += Inlines.SHR32(Inlines.MULT16_16(x, y), 2);
pyy += Inlines.SHR32(Inlines.MULT16_16(y, y), 2);
xx += Inlines.SHR32(pxx, 10);
xy += Inlines.SHR32(pxy, 10);
yy += Inlines.SHR32(pyy, 10);
}
mem.XX += Inlines.MULT16_32_Q15(short_alpha, xx - mem.XX);
mem.XY += Inlines.MULT16_32_Q15(short_alpha, xy - mem.XY);
mem.YY += Inlines.MULT16_32_Q15(short_alpha, yy - mem.YY);
mem.XX = Inlines.MAX32(0, mem.XX);
mem.XY = Inlines.MAX32(0, mem.XY);
mem.YY = Inlines.MAX32(0, mem.YY);
if (Inlines.MAX32(mem.XX, mem.YY) > ((short)(0.5 + (8e-4f) * (((int)1) << (18)))) /*Inlines.QCONST16(8e-4f, 18)*/)
{
sqrt_xx = Inlines.celt_sqrt(mem.XX);
sqrt_yy = Inlines.celt_sqrt(mem.YY);
qrrt_xx = Inlines.celt_sqrt(sqrt_xx);
qrrt_yy = Inlines.celt_sqrt(sqrt_yy);
/* Inter-channel correlation */
mem.XY = Inlines.MIN32(mem.XY, sqrt_xx * sqrt_yy);
corr = Inlines.SHR32(Inlines.frac_div32(mem.XY, CeltConstants.EPSILON + Inlines.MULT16_16(sqrt_xx, sqrt_yy)), 16);
/* Approximate loudness difference */
ldiff = CeltConstants.Q15ONE * Inlines.ABS16(qrrt_xx - qrrt_yy) / (CeltConstants.EPSILON + qrrt_xx + qrrt_yy);
width = Inlines.MULT16_16_Q15(Inlines.celt_sqrt(((int)(0.5 + (1.0f) * (((int)1) << (30)))) /*Inlines.QCONST32(1.0f, 30)*/ - Inlines.MULT16_16(corr, corr)), ldiff);
/* Smoothing over one second */
mem.smoothed_width += (width - mem.smoothed_width) / frame_rate;
/* Peak follower */
mem.max_follower = Inlines.MAX16(mem.max_follower - ((short)(0.5 + (.02f) * (((int)1) << (15)))) /*Inlines.QCONST16(.02f, 15)*/ / frame_rate, mem.smoothed_width);
}
else
{
width = 0;
corr = CeltConstants.Q15ONE;
ldiff = 0;
}
/*printf("%f %f %f %f %f ", corr/(float)1.0f, ldiff/(float)1.0f, width/(float)1.0f, mem.smoothed_width/(float)1.0f, mem.max_follower/(float)1.0f);*/
return(Inlines.EXTRACT16(Inlines.MIN32(CeltConstants.Q15ONE, 20 * mem.max_follower)));
}
internal static uint alg_quant(int[] X, int X_ptr, int N, int K, int spread, int B, EntropyCoder enc
)
{
int[] y = new int[N];
int[] iy = new int[N];
int[] signx = new int[N];
int i, j;
int s;
int pulsesLeft;
int sum;
int xy;
int yy;
uint collapse_mask;
Inlines.OpusAssert(K > 0, "alg_quant() needs at least one pulse");
Inlines.OpusAssert(N > 1, "alg_quant() needs at least two dimensions");
exp_rotation(X, X_ptr, N, 1, B, K, spread);
/* Get rid of the sign */
sum = 0;
j = 0;
do
{
int xpj = X_ptr + j;
/* OPT: Make sure the following two lines result in conditional moves
* rather than branches. */
signx[j] = X[xpj] > 0 ? 1 : -1;
X[xpj] = Inlines.ABS16(X[xpj]);
iy[j] = 0;
y[j] = 0;
} while (++j < N);
xy = yy = 0;
pulsesLeft = K;
/* Do a pre-search by projecting on the pyramid */
if (K > (N >> 1))
{
int rcp;
j = 0; do
{
sum += X[X_ptr + j];
} while (++j < N);
/* If X is too small, just replace it with a pulse at 0 */
/* Prevents infinities and NaNs from causing too many pulses
* to be allocated. 64 is an approximation of infinity here. */
if (sum <= K)
{
X[X_ptr] = ((short)(0.5 + (1.0f) * (((int)1) << (14)))) /*Inlines.QCONST16(1.0f, 14)*/;
j = X_ptr + 1;
do
{
X[j] = 0;
} while (++j < N + X_ptr);
sum = ((short)(0.5 + (1.0f) * (((int)1) << (14)))) /*Inlines.QCONST16(1.0f, 14)*/;
}
rcp = Inlines.EXTRACT16(Inlines.MULT16_32_Q16((K - 1), Inlines.celt_rcp(sum)));
j = 0;
do
{
/* It's really important to round *towards zero* here */
iy[j] = Inlines.MULT16_16_Q15(X[X_ptr + j], rcp);
y[j] = (int)iy[j];
yy = (Inlines.MAC16_16(yy, y[j], y[j]));
xy = Inlines.MAC16_16(xy, X[X_ptr + j], y[j]);
y[j] *= 2;
pulsesLeft -= iy[j];
} while (++j < N);
}
Inlines.OpusAssert(pulsesLeft >= 1, "Allocated too many pulses in the quick pass");
/* This should never happen, but just in case it does (e.g. on silence)
* we fill the first bin with pulses. */
if (pulsesLeft > N + 3)
{
int tmp = (int)pulsesLeft;
yy = (Inlines.MAC16_16(yy, tmp, tmp));
yy = (Inlines.MAC16_16(yy, tmp, y[0]));
iy[0] += pulsesLeft;
pulsesLeft = 0;
}
s = 1;
for (i = 0; i < pulsesLeft; i++)
{
int best_id;
int best_num = 0 - CeltConstants.VERY_LARGE16;
int best_den = 0;
int rshift = 1 + Inlines.celt_ilog2(K - pulsesLeft + i + 1);
best_id = 0;
/* The squared magnitude term gets added anyway, so we might as well
* add it outside the loop */
yy = Inlines.ADD16(yy, 1); // opus bug - was add32
j = 0;
do
{
int Rxy, Ryy;
/* Temporary sums of the new pulse(s) */
Rxy = Inlines.EXTRACT16(Inlines.SHR32(Inlines.ADD32(xy, Inlines.EXTEND32(X[X_ptr + j])), rshift));
/* We're multiplying y[j] by two so we don't have to do it here */
Ryy = Inlines.ADD16(yy, y[j]);
/* Approximate score: we maximise Rxy/sqrt(Ryy) (we're guaranteed that
* Rxy is positive because the sign is pre-computed) */
Rxy = Inlines.MULT16_16_Q15(Rxy, Rxy);
/* The idea is to check for num/den >= best_num/best_den, but that way
* we can do it without any division */
/* OPT: Make sure to use conditional moves here */
if (Inlines.MULT16_16(best_den, Rxy) > Inlines.MULT16_16(Ryy, best_num))
{
best_den = Ryy;
best_num = Rxy;
best_id = j;
}
} while (++j < N);
/* Updating the sums of the new pulse(s) */
xy = Inlines.ADD32(xy, Inlines.EXTEND32(X[X_ptr + best_id]));
/* We're multiplying y[j] by two so we don't have to do it here */
yy = Inlines.ADD16(yy, y[best_id]);
/* Only now that we've made the final choice, update y/iy */
/* Multiplying y[j] by 2 so we don't have to do it everywhere else */
y[best_id] = (y[best_id] + (2 * s));
iy[best_id]++;
}
/* Put the original sign back */
j = 0;
do
{
X[X_ptr + j] = (Inlines.MULT16_16(signx[j], X[X_ptr + j]));
/* OPT: Make sure your compiler uses a conditional move here rather than
* a branch. */
iy[j] = signx[j] < 0 ? -iy[j] : iy[j];
} while (++j < N);
CWRS.encode_pulses(iy, N, K, enc);
collapse_mask = extract_collapse_mask(iy, N, B);
return(collapse_mask);
}
/// <summary>
///
/// </summary>
/// <typeparam name="T">The type of signal being handled (either short or float) - changes based on which API is used</typeparam>
/// <param name="tonal"></param>
/// <param name="celt_mode"></param>
/// <param name="x"></param>
/// <param name="len"></param>
/// <param name="offset"></param>
/// <param name="c1"></param>
/// <param name="c2"></param>
/// <param name="C"></param>
/// <param name="lsb_depth"></param>
/// <param name="downmix"></param>
internal static void tonality_analysis <T>(TonalityAnalysisState tonal, CeltMode celt_mode, T[] x, int x_ptr, int len, int offset, int c1, int c2, int C, int lsb_depth, Downmix.downmix_func <T> downmix)
{
int i, b;
FFTState kfft;
int[] input;
int[] output;
int N = 480, N2 = 240;
float[] A = tonal.angle;
float[] dA = tonal.d_angle;
float[] d2A = tonal.d2_angle;
float[] tonality;
float[] noisiness;
float[] band_tonality = new float[OpusConstants.NB_TBANDS];
float[] logE = new float[OpusConstants.NB_TBANDS];
float[] BFCC = new float[8];
float[] features = new float[25];
float frame_tonality;
float max_frame_tonality;
/*float tw_sum=0;*/
float frame_noisiness;
float pi4 = (float)(M_PI * M_PI * M_PI * M_PI);
float slope = 0;
float frame_stationarity;
float relativeE;
float[] frame_probs = new float[2];
float alpha, alphaE, alphaE2;
float frame_loudness;
float bandwidth_mask;
int bandwidth = 0;
float maxE = 0;
float noise_floor;
int remaining;
AnalysisInfo info; //[porting note] pointer
tonal.last_transition++;
alpha = 1.0f / Inlines.IMIN(20, 1 + tonal.count);
alphaE = 1.0f / Inlines.IMIN(50, 1 + tonal.count);
alphaE2 = 1.0f / Inlines.IMIN(1000, 1 + tonal.count);
if (tonal.count < 4)
{
tonal.music_prob = 0.5f;
}
kfft = celt_mode.mdct.kfft[0];
if (tonal.count == 0)
{
tonal.mem_fill = 240;
}
downmix(x, x_ptr, tonal.inmem, tonal.mem_fill, Inlines.IMIN(len, OpusConstants.ANALYSIS_BUF_SIZE - tonal.mem_fill), offset, c1, c2, C);
if (tonal.mem_fill + len < OpusConstants.ANALYSIS_BUF_SIZE)
{
tonal.mem_fill += len;
/* Don't have enough to update the analysis */
return;
}
info = tonal.info[tonal.write_pos++];
if (tonal.write_pos >= OpusConstants.DETECT_SIZE)
{
tonal.write_pos -= OpusConstants.DETECT_SIZE;
}
input = new int[960];
output = new int[960];
tonality = new float[240];
noisiness = new float[240];
for (i = 0; i < N2; i++)
{
float w = Tables.analysis_window[i];
input[2 * i] = (int)(w * tonal.inmem[i]);
input[2 * i + 1] = (int)(w * tonal.inmem[N2 + i]);
input[(2 * (N - i - 1))] = (int)(w * tonal.inmem[N - i - 1]);
input[(2 * (N - i - 1)) + 1] = (int)(w * tonal.inmem[N + N2 - i - 1]);
}
Arrays.MemMoveInt(tonal.inmem, OpusConstants.ANALYSIS_BUF_SIZE - 240, 0, 240);
remaining = len - (OpusConstants.ANALYSIS_BUF_SIZE - tonal.mem_fill);
downmix(x, x_ptr, tonal.inmem, 240, remaining, offset + OpusConstants.ANALYSIS_BUF_SIZE - tonal.mem_fill, c1, c2, C);
tonal.mem_fill = 240 + remaining;
KissFFT.opus_fft(kfft, input, output);
for (i = 1; i < N2; i++)
{
float X1r, X2r, X1i, X2i;
float angle, d_angle, d2_angle;
float angle2, d_angle2, d2_angle2;
float mod1, mod2, avg_mod;
X1r = (float)output[2 * i] + output[2 * (N - i)];
X1i = (float)output[(2 * i) + 1] - output[2 * (N - i) + 1];
X2r = (float)output[(2 * i) + 1] + output[2 * (N - i) + 1];
X2i = (float)output[2 * (N - i)] - output[2 * i];
angle = (float)(.5f / M_PI) * fast_atan2f(X1i, X1r);
d_angle = angle - A[i];
d2_angle = d_angle - dA[i];
angle2 = (float)(.5f / M_PI) * fast_atan2f(X2i, X2r);
d_angle2 = angle2 - angle;
d2_angle2 = d_angle2 - d_angle;
mod1 = d2_angle - (float)Math.Floor(0.5f + d2_angle);
noisiness[i] = Inlines.ABS16(mod1);
mod1 *= mod1;
mod1 *= mod1;
mod2 = d2_angle2 - (float)Math.Floor(0.5f + d2_angle2);
noisiness[i] += Inlines.ABS16(mod2);
mod2 *= mod2;
mod2 *= mod2;
avg_mod = .25f * (d2A[i] + 2.0f * mod1 + mod2);
tonality[i] = 1.0f / (1.0f + 40.0f * 16.0f * pi4 * avg_mod) - .015f;
A[i] = angle2;
dA[i] = d_angle2;
d2A[i] = mod2;
}
frame_tonality = 0;
max_frame_tonality = 0;
/*tw_sum = 0;*/
info.activity = 0;
frame_noisiness = 0;
frame_stationarity = 0;
if (tonal.count == 0)
{
for (b = 0; b < OpusConstants.NB_TBANDS; b++)
{
tonal.lowE[b] = 1e10f;
tonal.highE[b] = -1e10f;
}
}
relativeE = 0;
frame_loudness = 0;
for (b = 0; b < OpusConstants.NB_TBANDS; b++)
{
float E = 0, tE = 0, nE = 0;
float L1, L2;
float stationarity;
for (i = Tables.tbands[b]; i < Tables.tbands[b + 1]; i++)
{
float binE = output[2 * i] * (float)output[2 * i] + output[2 * (N - i)] * (float)output[2 * (N - i)]
+ output[2 * i + 1] * (float)output[2 * i + 1] + output[2 * (N - i) + 1] * (float)output[2 * (N - i) + 1];
/* FIXME: It's probably best to change the BFCC filter initial state instead */
binE *= 5.55e-17f;
E += binE;
tE += binE * tonality[i];
nE += binE * 2.0f * (.5f - noisiness[i]);
}
tonal.E[tonal.E_count][b] = E;
frame_noisiness += nE / (1e-15f + E);
frame_loudness += (float)Math.Sqrt(E + 1e-10f);
logE[b] = (float)Math.Log(E + 1e-10f);
tonal.lowE[b] = Inlines.MIN32(logE[b], tonal.lowE[b] + 0.01f);
tonal.highE[b] = Inlines.MAX32(logE[b], tonal.highE[b] - 0.1f);
if (tonal.highE[b] < tonal.lowE[b] + 1.0f)
{
tonal.highE[b] += 0.5f;
tonal.lowE[b] -= 0.5f;
}
relativeE += (logE[b] - tonal.lowE[b]) / (1e-15f + tonal.highE[b] - tonal.lowE[b]);
L1 = L2 = 0;
for (i = 0; i < OpusConstants.NB_FRAMES; i++)
{
L1 += (float)Math.Sqrt(tonal.E[i][b]);
L2 += tonal.E[i][b];
}
stationarity = Inlines.MIN16(0.99f, L1 / (float)Math.Sqrt(1e-15 + OpusConstants.NB_FRAMES * L2));
stationarity *= stationarity;
stationarity *= stationarity;
frame_stationarity += stationarity;
/*band_tonality[b] = tE/(1e-15+E)*/
band_tonality[b] = Inlines.MAX16(tE / (1e-15f + E), stationarity * tonal.prev_band_tonality[b]);
frame_tonality += band_tonality[b];
if (b >= OpusConstants.NB_TBANDS - OpusConstants.NB_TONAL_SKIP_BANDS)
{
frame_tonality -= band_tonality[b - OpusConstants.NB_TBANDS + OpusConstants.NB_TONAL_SKIP_BANDS];
}
max_frame_tonality = Inlines.MAX16(max_frame_tonality, (1.0f + .03f * (b - OpusConstants.NB_TBANDS)) * frame_tonality);
slope += band_tonality[b] * (b - 8);
tonal.prev_band_tonality[b] = band_tonality[b];
}
bandwidth_mask = 0;
bandwidth = 0;
maxE = 0;
noise_floor = 5.7e-4f / (1 << (Inlines.IMAX(0, lsb_depth - 8)));
noise_floor *= 1 << (15 + CeltConstants.SIG_SHIFT);
noise_floor *= noise_floor;
for (b = 0; b < OpusConstants.NB_TOT_BANDS; b++)
{
float E = 0;
int band_start, band_end;
/* Keep a margin of 300 Hz for aliasing */
band_start = Tables.extra_bands[b];
band_end = Tables.extra_bands[b + 1];
for (i = band_start; i < band_end; i++)
{
float binE = output[2 * i] * (float)output[2 * i] + output[2 * (N - i)] * (float)output[2 * (N - i)]
+ output[2 * i + 1] * (float)output[2 * i + 1] + output[2 * (N - i) + 1] * (float)output[2 * (N - i) + 1];
E += binE;
}
maxE = Inlines.MAX32(maxE, E);
tonal.meanE[b] = Inlines.MAX32((1 - alphaE2) * tonal.meanE[b], E);
E = Inlines.MAX32(E, tonal.meanE[b]);
/* Use a simple follower with 13 dB/Bark slope for spreading function */
bandwidth_mask = Inlines.MAX32(.05f * bandwidth_mask, E);
/* Consider the band "active" only if all these conditions are met:
* 1) less than 10 dB below the simple follower
* 2) less than 90 dB below the peak band (maximal masking possible considering
* both the ATH and the loudness-dependent slope of the spreading function)
* 3) above the PCM quantization noise floor
*/
if (E > .1 * bandwidth_mask && E * 1e9f > maxE && E > noise_floor * (band_end - band_start))
{
bandwidth = b;
}
}
if (tonal.count <= 2)
{
bandwidth = 20;
}
frame_loudness = 20 * (float)Math.Log10(frame_loudness);
tonal.Etracker = Inlines.MAX32(tonal.Etracker - .03f, frame_loudness);
tonal.lowECount *= (1 - alphaE);
if (frame_loudness < tonal.Etracker - 30)
{
tonal.lowECount += alphaE;
}
for (i = 0; i < 8; i++)
{
float sum = 0;
for (b = 0; b < 16; b++)
{
sum += Tables.dct_table[i * 16 + b] * logE[b];
}
BFCC[i] = sum;
}
frame_stationarity /= OpusConstants.NB_TBANDS;
relativeE /= OpusConstants.NB_TBANDS;
if (tonal.count < 10)
{
relativeE = 0.5f;
}
frame_noisiness /= OpusConstants.NB_TBANDS;
info.activity = frame_noisiness + (1 - frame_noisiness) * relativeE;
frame_tonality = (max_frame_tonality / (OpusConstants.NB_TBANDS - OpusConstants.NB_TONAL_SKIP_BANDS));
frame_tonality = Inlines.MAX16(frame_tonality, tonal.prev_tonality * .8f);
tonal.prev_tonality = frame_tonality;
slope /= 8 * 8;
info.tonality_slope = slope;
tonal.E_count = (tonal.E_count + 1) % OpusConstants.NB_FRAMES;
tonal.count++;
info.tonality = frame_tonality;
for (i = 0; i < 4; i++)
{
features[i] = -0.12299f * (BFCC[i] + tonal.mem[i + 24]) + 0.49195f * (tonal.mem[i] + tonal.mem[i + 16]) + 0.69693f * tonal.mem[i + 8] - 1.4349f * tonal.cmean[i];
}
for (i = 0; i < 4; i++)
{
tonal.cmean[i] = (1 - alpha) * tonal.cmean[i] + alpha * BFCC[i];
}
for (i = 0; i < 4; i++)
{
features[4 + i] = 0.63246f * (BFCC[i] - tonal.mem[i + 24]) + 0.31623f * (tonal.mem[i] - tonal.mem[i + 16]);
}
for (i = 0; i < 3; i++)
{
features[8 + i] = 0.53452f * (BFCC[i] + tonal.mem[i + 24]) - 0.26726f * (tonal.mem[i] + tonal.mem[i + 16]) - 0.53452f * tonal.mem[i + 8];
}
if (tonal.count > 5)
{
for (i = 0; i < 9; i++)
{
tonal.std[i] = (1 - alpha) * tonal.std[i] + alpha * features[i] * features[i];
}
}
for (i = 0; i < 8; i++)
{
tonal.mem[i + 24] = tonal.mem[i + 16];
tonal.mem[i + 16] = tonal.mem[i + 8];
tonal.mem[i + 8] = tonal.mem[i];
tonal.mem[i] = BFCC[i];
}
for (i = 0; i < 9; i++)
{
features[11 + i] = (float)Math.Sqrt(tonal.std[i]);
}
features[20] = info.tonality;
features[21] = info.activity;
features[22] = frame_stationarity;
features[23] = info.tonality_slope;
features[24] = tonal.lowECount;
mlp.mlp_process(Tables.net, features, frame_probs);
frame_probs[0] = .5f * (frame_probs[0] + 1);
/* Curve fitting between the MLP probability and the actual probability */
frame_probs[0] = .01f + 1.21f * frame_probs[0] * frame_probs[0] - .23f * (float)Math.Pow(frame_probs[0], 10);
/* Probability of active audio (as opposed to silence) */
frame_probs[1] = .5f * frame_probs[1] + .5f;
/* Consider that silence has a 50-50 probability. */
frame_probs[0] = frame_probs[1] * frame_probs[0] + (1 - frame_probs[1]) * .5f;
/*printf("%f %f ", frame_probs[0], frame_probs[1]);*/
{
/* Probability of state transition */
float tau;
/* Represents independence of the MLP probabilities, where
* beta=1 means fully independent. */
float beta;
/* Denormalized probability of speech (p0) and music (p1) after update */
float p0, p1;
/* Probabilities for "all speech" and "all music" */
float s0, m0;
/* Probability sum for renormalisation */
float psum;
/* Instantaneous probability of speech and music, with beta pre-applied. */
float speech0;
float music0;
/* One transition every 3 minutes of active audio */
tau = .00005f * frame_probs[1];
beta = .05f;
//if (1)
{
/* Adapt beta based on how "unexpected" the new prob is */
float p, q;
p = Inlines.MAX16(.05f, Inlines.MIN16(.95f, frame_probs[0]));
q = Inlines.MAX16(.05f, Inlines.MIN16(.95f, tonal.music_prob));
beta = .01f + .05f * Inlines.ABS16(p - q) / (p * (1 - q) + q * (1 - p));
}
/* p0 and p1 are the probabilities of speech and music at this frame
* using only information from previous frame and applying the
* state transition model */
p0 = (1 - tonal.music_prob) * (1 - tau) + tonal.music_prob * tau;
p1 = tonal.music_prob * (1 - tau) + (1 - tonal.music_prob) * tau;
/* We apply the current probability with exponent beta to work around
* the fact that the probability estimates aren't independent. */
p0 *= (float)Math.Pow(1 - frame_probs[0], beta);
p1 *= (float)Math.Pow(frame_probs[0], beta);
/* Normalise the probabilities to get the Marokv probability of music. */
tonal.music_prob = p1 / (p0 + p1);
info.music_prob = tonal.music_prob;
/* This chunk of code deals with delayed decision. */
psum = 1e-20f;
/* Instantaneous probability of speech and music, with beta pre-applied. */
speech0 = (float)Math.Pow(1 - frame_probs[0], beta);
music0 = (float)Math.Pow(frame_probs[0], beta);
if (tonal.count == 1)
{
tonal.pspeech[0] = 0.5f;
tonal.pmusic[0] = 0.5f;
}
/* Updated probability of having only speech (s0) or only music (m0),
* before considering the new observation. */
s0 = tonal.pspeech[0] + tonal.pspeech[1];
m0 = tonal.pmusic[0] + tonal.pmusic[1];
/* Updates s0 and m0 with instantaneous probability. */
tonal.pspeech[0] = s0 * (1 - tau) * speech0;
tonal.pmusic[0] = m0 * (1 - tau) * music0;
/* Propagate the transition probabilities */
for (i = 1; i < OpusConstants.DETECT_SIZE - 1; i++)
{
tonal.pspeech[i] = tonal.pspeech[i + 1] * speech0;
tonal.pmusic[i] = tonal.pmusic[i + 1] * music0;
}
/* Probability that the latest frame is speech, when all the previous ones were music. */
tonal.pspeech[OpusConstants.DETECT_SIZE - 1] = m0 * tau * speech0;
/* Probability that the latest frame is music, when all the previous ones were speech. */
tonal.pmusic[OpusConstants.DETECT_SIZE - 1] = s0 * tau * music0;
/* Renormalise probabilities to 1 */
for (i = 0; i < OpusConstants.DETECT_SIZE; i++)
{
psum += tonal.pspeech[i] + tonal.pmusic[i];
}
psum = 1.0f / psum;
for (i = 0; i < OpusConstants.DETECT_SIZE; i++)
{
tonal.pspeech[i] *= psum;
tonal.pmusic[i] *= psum;
}
psum = tonal.pmusic[0];
for (i = 1; i < OpusConstants.DETECT_SIZE; i++)
{
psum += tonal.pspeech[i];
}
/* Estimate our confidence in the speech/music decisions */
if (frame_probs[1] > .75)
{
if (tonal.music_prob > .9)
{
float adapt;
adapt = 1.0f / (++tonal.music_confidence_count);
tonal.music_confidence_count = Inlines.IMIN(tonal.music_confidence_count, 500);
tonal.music_confidence += adapt * Inlines.MAX16(-.2f, frame_probs[0] - tonal.music_confidence);
}
if (tonal.music_prob < .1)
{
float adapt;
adapt = 1.0f / (++tonal.speech_confidence_count);
tonal.speech_confidence_count = Inlines.IMIN(tonal.speech_confidence_count, 500);
tonal.speech_confidence += adapt * Inlines.MIN16(.2f, frame_probs[0] - tonal.speech_confidence);
}
}
else
{
if (tonal.music_confidence_count == 0)
{
tonal.music_confidence = .9f;
}
if (tonal.speech_confidence_count == 0)
{
tonal.speech_confidence = .1f;
}
}
}
if (tonal.last_music != ((tonal.music_prob > .5f) ? 1 : 0))
{
tonal.last_transition = 0;
}
tonal.last_music = (tonal.music_prob > .5f) ? 1 : 0;
info.bandwidth = bandwidth;
info.noisiness = frame_noisiness;
info.valid = 1;
}