Resampling changes the sample rate. Downsample needs anti-alias filter; upsample needs interpolation.

What you're seeing

Common conversions: 16kHz speech → 48kHz audio output. 48kHz → 44.1kHz for CD. Each requires careful filtering: downsampling without an anti-alias filter introduces aliasing artifacts.

Linux: swresample (ffmpeg). Python: scipy.signal.resample_poly or librosa.resample. Quality depends on filter length (kaiser, sinc).

★ KEY TAKEAWAY
Up- or down-sampling needs careful filtering. Bad resampling adds aliasing artifacts.
▶ WHAT TO TRY
  • Switch from 16kHz to 48kHz — see more sample points.
  • Real resampling uses sinc-based or polyphase filters, not just linear interpolation.
← Back to Audio Engineering