Derivative Kernels for Noise Robust ASR

Recently there has been interest in combined gen- erative/discriminative classifiers. In these classifiers features for the discriminative models are derived from generative kernels. One advantage of using generative kernels is that systematic approaches exist how to introduce complex dependencies beyond conditional independence assumptions. Furthermore, by using generative kernels model-based compensation/adaptation tech- niques can be applied to make discriminative models robust to noise/speaker conditions. This paper extends previous work with combined generative/discriminative classifiers in several directions. First, it introduces derivative kernels based on context- dependent generative models. Second, it describes how derivative kernels can be incorporated in continuous discriminative models. Third, it addresses the issues associated with large number of classes and parameters when context-dependent models and high- dimensional features of derivative kernels are used. The approach is evaluated on two noise-corrupted tasks: small vocabulary AURORA 2 and medium-to-large vocabulary AURORA 4 task.

A variational perspective on noise-robust speech recognition

Model compensation methods for noise-robust speech recognition have shown good performance. Predictive linear transformations can approximate these methods to balance computational complexity and compensation accuracy. This paper examines both of these approaches from a variational perspective. Using a matched-pair approximation at the component level yields a number of standard forms of model compensation and predictive linear transformations. However, a tighter bound can be obtained by using variational approximations at the state level. Both model-based and predictive linear transform schemes can be implemented in this framework. Preliminary results show that the tighter bound obtained from the state-level variational approach can yield improved performance over standard schemes. © 2011 IEEE.

Derivative kernels for noise robust ASR

Recently there has been interest in combining generative and discriminative classifiers. In these classifiers features for the discriminative models are derived from the generative kernels. One advantage of using generative kernels is that systematic approaches exist to introduce complex dependencies into the feature-space. Furthermore, as the features are based on generative models standard model-based compensation and adaptation techniques can be applied to make discriminative models robust to noise and speaker conditions. This paper extends previous work in this framework in several directions. First, it introduces derivative kernels based on context-dependent generative models. Second, it describes how derivative kernels can be incorporated in structured discriminative models. Third, it addresses the issues associated with large number of classes and parameters when context-dependent models and high-dimensional feature-spaces of derivative kernels are used. The approach is evaluated on two noise-corrupted tasks: small vocabulary AURORA 2 and medium-to-large vocabulary AURORA 4 task. © 2011 IEEE.

Extending noise robust structured support vector machines to larger vocabulary tasks

This paper describes a structured SVM framework suitable for noise-robust medium/large vocabulary speech recognition. Several theoretical and practical extensions to previous work on small vocabulary tasks are detailed. The joint feature space based on word models is extended to allow context-dependent triphone models to be used. By interpreting the structured SVM as a large margin log-linear model, illustrates that there is an implicit assumption that the prior of the discriminative parameter is a zero mean Gaussian. However, depending on the definition of likelihood feature space, a non-zero prior may be more appropriate. A general Gaussian prior is incorporated into the large margin training criterion in a form that allows the cutting plan algorithm to be directly applied. To further speed up the training process, 1-slack algorithm, caching competing hypothesis and parallelization strategies are also proposed. The performance of structured SVMs is evaluated on noise corrupted medium vocabulary speech recognition task: AURORA 4. © 2011 IEEE.

Gaussian Process training with input noise

In standard Gaussian Process regression input locations are assumed to be noise free. We present a simple yet effective GP model for training on input points corrupted by i.i.d. Gaussian noise. To make computations tractable we use a local linear expansion about each input point. This allows the input noise to be recast as output noise proportional to the squared gradient of the GP posterior mean. The input noise variances are inferred from the data as extra hyperparameters. They are trained alongside other hyperparameters by the usual method of maximisation of the marginal likelihood. Training uses an iterative scheme, which alternates between optimising the hyperparameters and calculating the posterior gradient. Analytic predictive moments can then be found for Gaussian distributed test points. We compare our model to others over a range of different regression problems and show that it improves over current methods.

Dual gate photo-thin film transistor with high photoconductive gain for high reliability, and low noise flat panel transparent imager

In this presentation, we report excellent electrical and optical characteristics of a dual gate photo thin film transistor (TFT) with bi-layer oxide channel, which was designed to provide virgin threshold voltage (V T) control, improve the negative bias illumination temperature stress (NBITS) reliability, and offer high photoconductive gain. In order to address the photo-sensitivity of phototransistor for the incoming light, top transparent InZnO (IZO) gate was employed, which enables the independent gate control of dual gate photo-TFT without having any degradation of its photosensitivity. Considering optimum initial V T and NBITS reliability for the device operation, the top gate bias was judiciously chosen. In addition, the speed and noise performance of the photo-TFT is competitive with silicon photo-transistors, and more importantly, its superiority lies in optical transparency. © 2011 IEEE.

Contra-rotating open rotor operation for improved aerodynamics and noise at takeoff

The contra-rotating open rotor is, once again, being considered as an alternative to the advanced turbofan to address the growing pressure to cut aviation fuel consumption and carbon dioxide emissions. One of the key challenges is meeting community noise targets at takeoff. Previous open rotor designs are subject to poor efficiency at takeoff due to the presence of large regions of separated flow on the blades as a result of the high incidence needed to achieve the required thrust. This is a consequence of the fixed rotor rotational speed constraint typical of variable pitch propellers. Within the study described in this paper, an improved operation is proposed to improve performance and reduce rotorrotor interaction noise at takeoff. Three-dimensional computational fluid dynamics (CFD) calculations have been performed on an open rotor rig at a range of takeoff operating conditions. These have been complemented by analytical tone noise predictions to quantify the noise benefits of the approach. The results presented show that for a given thrust, a combination of reduced rotor pitch and increased rotor rotational speed can be used to reduce the incidence onto the front rotor blades. This is shown to eliminate regions of flow separation, reduce the front rotor tip loss and reduce the downstream stream tube contraction. The wakes from the front rotor are also made wider with lower velocity defect, which is found to lead to reduced interaction tone noise. Unfortunately, the necessary increase in blade speed leads to higher relative Mach numbers, which can increase rotor alone noise. In summary, the combined CFD and aero-acoustic analysis in this paper shows how careful operation of an open rotor at takeoff, with moderate levels of re-pitch and speed increase, can lead to improved front rotor efficiency as well as appreciably lower overall noise across all directivities. Copyright © 2011 by ASME.

A design strategy in the propulsion system attachment to a submarine hull to minimise radiated noise

Vibration modes of a submerged hull are excited by fluctuating forces generated at the propeller and transmitted to the hull via the propeller-shafting system. The low frequency hull vibrational modes result in significant sound radiation. This work investigates the reduction of the far-field radiated sound pressure by optimising the connection point of the shafting system to the hull. The submarine hull is modelled as a fluid loaded cylindrical hull with truncated conical shells at each end. The propeller-shafting system consists of the propeller, shaft, thrust bearing and foundation, and is modelled in a modular approach using a combination of spring-mass-damper elements and continuous systems (beams, plates, shells). The foundation is attached to the stern side end plate of the hull, which is modelled as a circular plate coupled to an annular plate. By tuning the connection radius of the foundation to the end plate, the maximum radiated noise in a given frequency range can be minimised.

Shock Propagation and MPT Noise From a Transonic Rotor in Nonuniform Flow

One of the major challenges in high-speed fan stages used in compact, embedded propulsion systems is inlet distortion noise. A body-force-based approach for the prediction of multiple-pure-tone (MPT) noise was previously introduced and validated. In this paper, it is employed with the objective of quantifying the effects of nonuniform flow on the generation and propagation of MPT noise. First-of-their-kind back-to-back coupled aero-acoustic computations were carried out using the new approach for conventional and serpentine inlets. Both inlets delivered flow to the same NASA/GE R4 fan rotor at equal corrected mass flow rates. Although the source strength at the fan is increased by 38 dB in sound power level due to the nonuniform inflow, far-field noise for the serpentine inlet duct is increased on average by only 3.1 dBA overall sound pressure level in the forward arc. This is due to the redistribution of acoustic energy to frequencies below 11 times the shaft frequency and the apparent cut-off of tones at higher frequencies including blade-passing tones. The circumferential extent of the inlet swirl distortion at the fan was found to be two blade pitches, or 1/11th of the circumference, suggesting a relationship between the circumferential extent of the inlet distortion and the apparent cut-off frequency perceived in the far field. A first-principles-based model of the generation of shock waves from a transonic rotor in nonuniform flow showed that the effects of nonuniform flow on acoustic wave propagation, which cannot be captured by the simplified model, are more dominant than those of inlet flow distortion on source noise. It demonstrated that nonlinear, coupled aerodynamic and aero-acoustic computations, such as those presented in this paper, are necessary to assess the propagation through nonuniform mean flow. A parametric study of serpentine inlet designs is underway to quantify these propagation effects. © 2013 American Society of Mechanical Engineers.

Shock propagation and MPT noise from a transonic rotor in non-uniform flow

One of the major challenges in hig4h-speed fan stages used in compact, embedded propulsion systems is inlet distortion noise. A body-force-based approach for the prediction of multiple-pure-tone (MPT) noise was previously introduced and validated. In this paper, it is employed with the objective of quantifying the effects of non-uniform flow on the generation and propagation of MPT noise. First-of-their-kind back-to-back coupled aero-acoustic computations were carried out using the new approach for conventional and serpentine inlets. Both inlets delivered flow to the same NASA/GE R4 fan rotor at equal corrected mass flow rates. Although the source strength at the fan is increased by 45 dB in sound power level due to the non-uniform inflow, farfield noise for the serpentine inlet duct is increased on average by only 3.1 dBA overall sound pressure level in the forward arc. This is due to the redistribution of acoustic energy to frequencies below 11 times the shaft frequency and the apparent cut-off of tones at higher frequencies including blade-passing tones. The circumferential extent of the inlet swirl distortion at the fan was found to be 2 blade pitches, or 1/11th of the circumference, suggesting a relationship between the circumferential extent of the inlet distortion and the apparent cut-off frequency perceived in the far field. A first-principles-based model of the generation of shock waves from a transonic rotor in non-uniform flow showed that the effects of non-uniform flow on acoustic wave propagation, which cannot be captured by the simplified model, are more dominant than those of inlet flow distortion on source noise. It demonstrated that non-linear, coupled aerodynamic and aeroacoustic computations, such as those presented in this paper, are necessary to assess the propagation through non-uniform mean flow. A parametric study of serpentine inlet designs is underway to quantify these propagation effects. Copyright © 2011 by ASME.

A body force-based method for prediction of multiple-pure-tone noise: Validation

Emissions, fuel burn, and noise are the main drivers for innovative aircraft design. Embedded propulsion systems, such as for example used in hybrid-wing body aircraft, can offer fuel burn and noise reduction benefits but the impact of inlet flow distortion on the generation and propagation of turbomachinery noise has yet to be assessed. A novel approach is used to quantify the effects of non-uniform flow on the creation and propagation of multiple pure tone (MPT) noise. The ultimate goal is to conduct a parametric study of S-duct inlets to quantify the effects of inlet design parameters on the acoustic signature. The key challenge is that the effects of distortion transfer, noise source generation and propagation through the non-uniform flow field are inherently coupled such that a simultaneous computation of the aerodynamics and acoustics is required to capture the mechanisms at play. The technical approach is based on a body force description of the fan blade row that is able to capture the distortion transfer and the blade-to-blade flow variations that cause the MPT noise while reducing computational cost. A single, 3-D full-wheel CFD simulation, in which the Euler equations are solved to second-order spatial and temporal accuracy, simultaneously computes the MPT noise generation and its propagation in distorted inlet flow. A new method of producing the blade-to-blade variations in the body force field for MPT noise generation has been developed and validated. The numerical dissipation inherent to the solver is quantified and used to correct for non-physical attenuation in the far-field noise spectra. Source generation, acoustic propagation and acoustic energy transfer between modes is examined in detail. The new method is validated on NASA's Source Diagnostic Test fan and inlet, showing good agreement with experimental data for aerodynamic performance, acoustic source generation, and far-field noise spectra. The next steps involve the assessment of MPT noise in serpentine inlet ducts and the development of a reduced order formulation suitable for incorporation into NASA's ANOPP framework. © 2010 by Jeff Defoe, Alex Narkaj & Zoltan Spakovszky.