Programmiererforum

10*log10(fp^2/(fp-FP)^2))

print(fp .shape) #(100,300,125)
print(FP.shape) #(100,300,125)
for i in range(100) :
#1. SQNR[i] = 10 * math.log10(((fp[i]**2).sum())/((fp[i]-FP[i])**2).sum())
#2. SQNR[i] = 10 * math.log10(((np.mean(fp[i])**2))/(np.mean(fp[i]-FP[i])**2))
#3. SQNR[i] = 10 * math.log10(np.mean(fp[i]**2)/(np.mean(fp[i]-FP[i])**2))

print(np.mean(SQNR))

Die Formel für SQNR (Verhältnis von Signal-zu-Quantisation-Noise) lautet:
[code]10*log10(fp^2/(fp-FP)^2))[/code]
FP repräsentiert das Signal für Gleitpunkt (Original).[code]print(fp .shape) #(100,300,125)
print(FP.shape) #(100,300,125)
for i in range(100) :
#1. SQNR[i] = 10 * math.log10(((fp[i]**2).sum())/((fp[i]-FP[i])**2).sum())
#2. SQNR[i] = 10 * math.log10(((np.mean(fp[i])**2))/(np.mean(fp[i]-FP[i])**2))
#3. SQNR[i] = 10 * math.log10(np.mean(fp[i]**2)/(np.mean(fp[i]-FP[i])**2))

print(np.mean(SQNR))
[/code]