Dynamic Control of Voice Codec Data Rate

Real-time voice codec adaptation to devote more bits to important speech. Adjusting the bitrate based on inferred value of each frame. Frame importance is predicted based on models of dialog dynamics. No assumption that all speech is equally valuable. Current voice codes adapt to channel characteristics, but lack any adaptation to importance, communicative value, or dialog/reverence. Speech importance increases from less important silence, in-breaths, fillers (“Um”, “Well”), back channels (“Uh-huh”), disfluencies, false starts, and repetitions, to more important normal and critical content. Patent issued. Chao Zhang czhang3@utep.edu 915-747-6717