Hands-free terminals for voice over IP

Designing a hands-free terminal for voice over Internet protocol (VoIP) presents many challenges but also offers designers the opportunity to offer new features not currently available or even feasible on plain old telephone system (POTS) networks. This thesis focuses on techniques for offering hands-free spatial audio to the end-user and the problems associated with offering this enhanced feature set.;Multimedia-ready VoIP terminals are generally capable of offering stereo audio and some advanced terminals are even capable of offering surround sound audio to the end-user. Techniques for providing spatial audio to the near-end user are reviewed in this thesis and a recommended spatialization technique involving multiple speakers is made with reference to a common application programmer interface (API) called Microsoft DirectX.;VoIP systems typically suffer from long delays, which necessitate the addition of acoustic echo suppression techniques such as adaptive echo cancellers and optionally microphone array beamformers to hands-free terminals. This thesis proposes novel structures for synthetic stereo and synthetic surround sound audio presentation combined with acoustic echo cancellation. To further enhance echo suppression and provide increased near-end gain, this thesis proposes a combination of near-field microphone array beamforming and the synthetic stereo acoustic echo cancellation structure.;Long delays in a VoIP conference contribute to listener confusion and also require very large levels of echo suppression. It is therefore desirable to reduce the end-to-end delay of VoIP conferences to a minimum. A technique that compromises between the end-to-end delay and packet loss rate is proposed in this thesis that uses hidden Markov model (HMM) techniques. The proposed technique differs from other delay computation techniques since generic model parameters are precomputed in advance and are then applied to predict suitable VoIP conference delays.;The HMM technique or another existing playout delay computation technique can be used to compute the end-to-end delay as perceived by the user, called the acoustic round trip delay (ARTD). ARTD can be used to compute appropriate acoustic echo suppression targets using talker echo loudness rating curves, which indicate the appropriate amount of echo suppression required for a given end-to-end delay. This thesis proposes an echo suppression system for VoIP terminals that uses attenuators that are precisely controlled by echo suppression target levels to reduce half-duplex attenuation.;This thesis proposes several new technologies and shows unique ways of applying various existing technologies to provide enhanced hands-free full-duplex VoIP conferencing to the user. By following the recommendations made in this thesis, VoIP terminals will be able to provide users with an enhanced conferencing experience with new features previously infeasible on a POTS network.