A narrowband low bit rate sinusoidal audio and speech coder

Speech and audio coders use different coding strategies. As a result, broadband (22.05 kHz bandwidth) audio coders typically have a good quality for both audio and speech signals and a high bit rate, whilst narrowband (4 kHz bandwidth) speech coders have a low bit rate but a distorted quality for audio signals. Our aim is to develop a single narrowband low bit rate audio and speech coder that provides speech quality on par with current state-of-the-art speech coders and good audio quality. To this end, the coder uses a mix of typical speech and audio coding strategies. A prototype coder was implemented and compared to the GSM-EFR standard speech coder. For speech signals, the attained quality approaches that of the GSM-EFR coder, whilst for music it performs clearly better. These results raise the expectation that more transparent coders are feasible for future mobile telephony.