Standard Aurora Research Articles

Joint Uncertainty Decoding for Noise Robust Subspace Gaussian Mixture Models

Joint uncertainty decoding (JUD) is a model-based noise compensation technique for conventional Gaussian Mixture Model (GMM) based speech recognition systems. Unlike vector Taylor series (VTS) compensation which operates on the individual Gaussian components in an acoustic model, JUD clusters the Gaussian components into a smaller number of classes, sharing the compensation parameters for the set of Gaussians in a given class. This significantly reduces the computational cost. In this paper, we investigate noise compensation for subspace Gaussian mixture model (SGMM) based speech recognition systems using JUD. The total number of Gaussian components in an SGMM is typically very large. Therefore direct compensation of the individual Gaussian components, as performed by VTS, is computationally expensive. In this paper we show that JUD-based noise compensation can be successfully applied to SGMMs in a computationally efficient way. We evaluate the JUD/SGMM technique on the standard Aurora 4 corpus. Our experimental results indicate that the JUD/SGMM system results in lower word error rates compared with a conventional GMM system with either VTS-based or JUD-based noise compensation.

Standardized assessment of new electromagnetic field generators in an interventional radiology setting

Two of the main challenges associated with electromagnetic (EM) tracking in computer-assisted interventions (CAIs) are (1) the compensation of systematic distance errors arising from the influence of metal near the field generator (FG) or the tracked sensor and (2) the optimized setup of the FG to maximize tracking accuracy in the area of interest. Recently, two new FGs addressing these issues were proposed for the well-established Aurora(®) tracking system [Northern Digital, Inc. (NDI), Waterloo, Canada]: the Tabletop 50-70 FG, a planar transmitter with a built-in shield that compensates for metal distortions emanating from treatment tables, and the prototypical Compact FG 7-10, a mobile generator designed to be attached to mobile imaging devices. The purpose of this paper was to assess the accuracy and precision of these new FGs in an interventional radiology setting. A standardized assessment protocol, which uses a precisely machined base plate to measure relative error in position and orientation, was applied to the two new FGs as well as to the well-established standard Aurora(®) Planar FG. The experiments were performed in two different settings: a reference laboratory environment and a computed tomography (CT) scanning room. In each setting, the protocol was applied to three different poses of the measurement plate within the tracking volume of the three FGs. The two new FGs provided higher precision and accuracy within their respective measurement volumes as well as higher robustness with respect to the CT scanner compared to the established FG. Considering all possible 5 cm distances on the grid, the error of the Planar FG was increased by a factor of 5.94 in the clinical environment (4.4 mm) in comparison to the error in the laboratory environment (0.8 mm). In contrast, the mean values for the two new FGs were all below 1 mm with an increase in the error by factors of only 2.94 (Reference: 0.3 mm; CT: 0.9 mm) and 1.04 (both: 0.5 mm) in the case of the Tabletop FG and the Compact FG, respectively. Due to their high accuracy and robustness, the Tabletop FG and the Compact FG could eliminate the need for compensation of EM field distortions in certain CT-guided interventions.

A unified framework of HMM adaptation with joint compensation of additive and convolutive distortions

In this paper, we present our recent development of a model-domain environment robust adaptation algorithm, which demonstrates high performance in the standard Aurora 2 speech recognition task. The algorithm consists of two main steps. First, the noise and channel parameters are estimated using multi-sources of information including a nonlinear environment-distortion model in the cepstral domain, the posterior probabilities of all the Gaussians in speech recognizer, and truncated vector Taylor series (VTS) approximation. Second, the estimated noise and channel parameters are used to adapt the static and dynamic portions (delta and delta–delta) of the HMM means and variances. This two-step algorithm enables joint compensation of both additive and convolutive distortions (JAC). The hallmark of our new approach is the use of a nonlinear, phase-sensitive model of acoustic distortion that captures phase asynchrony between clean speech and the mixing noise. In the experimental evaluation using the standard Aurora 2 task, the proposed Phase-JAC/VTS algorithm achieves 93.32% word accuracy using the clean-trained complex HMM backend as the baseline system for the unsupervised model adaptation. This represents high recognition performance on this task without discriminative training of the HMM system. The experimental results show that the phase term, which was missing in all previous HMM adaptation work, contributes significantly to the achieved high recognition accuracy.