Concept: Quantization (signal processing)

Lower bit depth reduced speaker recognition accuracy

Lower-bit floating-point formats reduce memory while maintaining speaker identification performance