Spectral Subtraction for Speech Enhancement

Load speech file, generate random noise, synthesize noisy speech signal, and define parameter settings. Step 2: Frame segmentation with 50% overlap. Step 3: Apply Hamming window and perform Fourier transform to obtain power spectrum and phase spectrum. Step 4: Execute magnitude spectral subtraction and use the noisy speech phase for signal resynthesis via inverse Fourier transform. Step 5: Remove Hamming window to obtain enhanced speech signal. Step 6: Calculate SNR before and after enhancement.