The ARX-LF model interprets voiced speech as the an LF derivative glottal pulse exciting an all-pole vocal tract filter with an additional exogenous residual signal. It fully parameterizes the voice and has been shown to be useful for voice modification. Because time domain methods to determine the ARX-LF parameters from speech are very sensitive to the time placement of the analysis frame and not robust to phase distortion from e.g. recording equipment, a magnitude-only spectral approach to ARX-LF parameterization was recently developed. This paper describes extensions to this frequency domain approach to obtain continuous robust ARX-LF parameters for voiced speech segments. A listening test of 50 participants comparing synthetic speech produced by this method with a time domain ARX-LF parameterization approach under real and simulated recording conditions was conducted and it was found that the frequency domain approach was generally preferred.

It is crucial for many methods of inverse filtering that the time domain information of the glottal source waveform is known, e.g. the location of the instant of glottal closure. It is often the case that this information is unknown and/or cannot be determined due to e.g. recording conditions which can corrupt the phase spectrum. In these scenarios, alternative strategies are required. This paper describes a method which, given the parameters of the glottal formant of the signal frame, can accurately parameterize the glottal shape source and vocal filter for a broad range of voice quality types and which is robust to the corruption of the phase spectrum. Index Terms: glottal inverse filtering, frequency domain, glottal models, glottal formant

A framework is presented which addresses the issues related to the real-time implementation of synchronized video and audio time-scale and pitch-scale modification algorithms. It allows for seamless real-time transition between continually varying, independent time-scale and pitch-scale parameters arising as a result of manual or automatic intervention. We illuminate the problems which arise in a real-time context as well as provide novel solutions to prevent artifacts, minimize latency, and improve synchronization. The time and pitch scaling approach is based on a modified phase vocoder with optional phase locking and an integrated transient detector which enables high-quality transient preservation in real-time. A novel method for audio/visual synchronization was implemented in order to ensure no perceptible latency between audio and video while real-time time scaling and pitch shifting is applied. Evaluation results are reported which demonstrate both high audio quality and minimal synchronization error.

Acoustic/prosodic feature (a/p) convergence has been known to occur both in dialogues between humans, as well as in human-computer interactions. Understanding the form and function of convergence is desirable for developing next generation conversational agents, as this will help increase speech recognition performance and naturalness of synthesized speech. Currently, the underlying mechanisms by which continuous and bi-directional convergence occurs are not well understood. In this study, a direct comparison between time-aligned frames shows significant similarity in acoustic feature variation between the two speakers. The method described (TAMA) constitutes a first step towards a quantitative analysis of a/p convergence.

Time-domain audio time-scaling algorithms are efficient in comparison to their frequency-domain counterparts, but they rely upon the existence of a quasi-periodic signal to produce a high quality output. This requirement makes them unsuitable for direct application to complex multi-pitched signals such as polyphonic music. However, it has been shown that applying time-domain algorithms on a subband basis can resolve this issue. Existing subband/time-domain approaches result in a reverberant/phasy artifact being introduced into the output due to poor synchronization between time-scaled subbands. This paper presents a number of synchronization schemes that greatly reduce the amount of reverberation/phasiness introduced into the time-scaled output by existing subband/time-domain approaches.

The duration of a speech passage can be altered using audio time-scale modification techniques. Time-scale modification can be achieved in the time domain by segmenting the input signal into overlapping frames and recombining the frames with an overlap differing from the analysis overlap. We present a time-scale modification algorithm that uses a simple peak alignment technique to synchronize overlapping synthesis frames. The peak alignment overlap-add (PAOLA) algorithm also takes advantage of waveform properties to ensure a high quality output for the minimum number of iterations. The new algorithm produces a time-scaled output of approximately equal quality to that of an adaptive implementation of the commercially popular synchronised overlap-add (SOLA) algorithm, but offers a computational saving ranging from a factor of 15 (for a time-scale factor of 0.5) to 170 (for a time-scale factor of 1.1).