[Python] How to calculate SQNR?

The formula for SQNR (Signal-to-Quantization-Noise Ratio) is: 10*log10(fp^2/(fp-FP)^2))

fp represents the floating-point (original) signal.

FP represents the quantized (fixed-point) signal.

I don't know which one corresponds correctly in the code

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

print(np.mean(SQNR))

Continue reading...