Kaiserslautern - Fachbereich Elektrotechnik und Informationstechnik
Refine
Year of publication
- 2014 (3) (remove)
Document Type
- Doctoral Thesis (3) (remove)
Language
- English (3)
Has Fulltext
- yes (3)
Keywords
- Feature extraction (1)
- Noise control (1)
- Speech recognition (1)
Faculty / Organisational entity
The objective of this thesis consists in developing systematic event-triggered control designs for specified event generators, which is an important alternative to the traditional periodic sampling control. Sporadic sampling inherently arising in event-triggered control is determined by the event-triggering conditions. This feature invokes the desire of
finding new control theory as the traditional sampled-data theory in computer control.
Developing controller coupling with the applied event-triggering condition to maximize the control performance is the essence for event-triggered control design. In the design the stability of the control system needs to be ensured with the first priority. Concerning variant control aims they should be clearly incorporated in the design procedures. Considering applications in embedded control systems efficient implementation requires a low complexity of embedded software architectures. The thesis targets at offering such a design to further complete the theory of event-triggered control designs.
In this thesis we studied and investigated a very common but a long existing noise problem and we provided a solution to this problem. The task is to deal with different types of noise that occur simultaneously and which we call hybrid. Although there are individual solutions for specific types one cannot simply combine them because each solution affects the whole speech. We developed an automatic speech recognition system DANSR ( Dynamic Automatic Noisy Speech Recognition System) for hybrid noisy environmental noise. For this we had to study all of speech starting from the production of sounds until their recognition. Central elements are the feature vectors on which pay much attention. As an additional effect we worked on the production of quantities for psychoacoustic speech elements.
The thesis has four parts:
1) The first part we give an introduction. The chapter 2 and 3 give an overview over speech generation and recognition when machines are used. Also noise is considered.
2) In the second part we describe our general system for speech recognition in a noisy environment. This is contained in the chapters 4-10. In chapter 4 we deal with data preparation. Chapter 5 is concerned with very strong noise and its modeling using Poisson distribution. In the chapters 5-8 we deal with parameter based modeling. Chapter 7 is concerned with autoregressive methods in relation to the vocal tract. In the chapters 8 and 9 we discuss linear prediction and its parameters. Chapter 9 is also concerned with quadratic errors, the decomposition into sub-bands and the use of Kalman filters for non-stationary colored noise in chapter 10. There one finds classical approaches as long we have used and modified them. This includes covariance mehods, the method of Burg and others.
3) The third part deals firstly with psychoacoustic questions. We look at quantitative magnitudes that describe them. This has serious consequences for the perception models. For hearing we use different scales and filters. In the center of the chapters 12 and 13 one finds the features and their extraction. The fearures are the only elements that contain information for further use. We consider here Cepstrum features and Mel frequency cepstral coefficients(MFCC), shift invariant local trigonometric transformed (SILTT), linear predictive coefficients (LPC), linear predictive cepstral coefficients (LPCC), perceptual linear predictive (PLP) cepstral coefficients. In chapter 13 we present our extraction methods in DANSR and how they use window techniques And discrete cosine transform (DCT-IV) as well as their inverses.
4) The fourth part considers classification and the ultimate speech recognition. Here we use the hidden Markov model (HMM) for describing the speech process and the Gaussian mixture model (GMM) for the acoustic modelling. For the recognition we use forward algorithm, the Viterbi search and the Baum-Welch algorithm. We also draw the connection to dynamic time warping (DTW). In the rest we show experimental results and conclusions.
The work presented in this thesis discusses the thermal and power management of multi-core processors (MCPs) with both two dimensional (2D) package and there dimensional (3D) package chips. The power and thermal management/balancing is of increasing concern and is a technological challenge to the MCP development and will be a main performance bottleneck for the development of MCPs. This thesis develops optimal thermal and power management policies for MCPs. The system thermal behavior for both 2D package and 3D package chips is analyzed and mathematical models are developed. Thereafter, the optimal thermal and power management methods are introduced.
Nowadays, the chips are generally packed in 2D technique, which means that there is only one layer of dies in the chip. The chip thermal behavior can be described by a 3D heat conduction partial differential equation (PDE). As the target is to balance the thermal behavior and power consumption among the cores, a group of one dimensional (1D) PDEs, which is derived from the developed 3D PDE heat conduction equation, is proposed to describe the thermal behavior of each core. Therefore, the thermal behavior of the MCP is described by a group of 1D PDEs. An optimal controller is designed to manage the power consumption and balance the temperature among the cores based on the proposed 1D model.
3D package is an advanced package technology, which contains at least 2 layers of dies stacked in one chip. Different from 2D package, the cooling system should be installed among the layers to reduce the internal temperature of the chip. In this thesis, the micro-channel liquid cooling system is considered, and the heat transfer character of the micro-channel is analyzed and modeled as an ordinary differential equation (ODE). The dies are discretized to blocks based on the chip layout with each block modeled as a thermal resistance and capacitance (R-C) circuit. Thereafter, the micro-channels are discretized. The thermal behavior of the whole system is modeled as an ODE system. The micro-channel liquid velocity is set according to the workload and the temperature of the dies. Under each velocity, the system can be described as a linear ODE model system and the whole system is a switched linear system. An H-infinity observer is designed to estimate the states. The model predictive control (MPC) method is employed to design the thermal and power management/balancing controller for each submodel.
The models and controllers developed in this thesis are verified by simulation experiments via MATLAB. The IBM cell 8 cores processor and water micro-channel cooling system developed by IBM Research in collaboration with EPFL and ETHZ are employed as the experiment objects.