Our team is developing an integrated algorithm analysis and mapping environment for migrating a dataflow representation of a signal processing algorithm into an Adaptive Computing System (ACS) consisting of FPGAs. This environment allows designers to transform signal processing algorithms into FPGA-based hardware faster, by an order of magnitude, than is currently possible. Our approach has been to focus on three areas of capability critical to the success of adaptive computing: algorithm analysis, algorithm mapping, and smart generators. These capabilities are being implemented as extensions to the Ptolemy design environment developed at the University of California, Berkeley (http://ptolemy.cecs.berkeley.edu). Our tools were used to automatically implement a Winograd DFT and a high speed linear FM detector. In both cases, our tools simulated the algorithm, selected appropriate fixed point representations, and generated the VHDL implementations. The final FPGA designs were obtained by synthesizing the VHDL and performing place and route with commercial tools.

This paper presents modeling approaches and experiences in the use of the VHSIC hardware description language (VHDL) for the development of application-specific signal processors. Within our work on the ARPA/Tri-Service RASSP program we have developed and used VHDL modeling techniques for modeling the performance of signal processor systems and for the detailed design of signal processor systems. These approaches have been applied to modeling a large infra-red search and track system from a functional model, through performance modeling, through a full functional model, down to a detailed hardware implementation model.

This paper presents two case studies of the rapid development of military signal processors and a description of how the experiences learned on these programs are being leveraged into an ARPA/Tri-Service initiative. The two case studies emphasize end-to-end system simulation and the use of rapid prototyping tools and special hardware for rapid development. Rapid Prototyping of Application Specific Signal Processors, or RASSP, is an initiative to create a new process for the development of military signal processors. The objective of RASSP is to specify a process which both reduces the development time and the cost of military signal processing systems while improving their quality and their long term supportability. RASSP processors are meant to be incrementally upgradable over long development and life cycles. While RASSP represents a significantly harder problem than the two examples described here, the lessons learned in these examples are being incorporated in the RASSP approach.>

The problem of signal separation for nonlinear dynamical systems, particularly chaotic systems, is considered. These systems are characterized by a stretching and folding within state space and by the presence of an attractor. Signal separation involves the separation of a received signal into two components, one of which is modeled as the output of a nonlinear dynamical system. The authors review previous approaches to this problem and present results from the application of Kalman filtering to the signal separation problem. A Cramer-Rao bound on the performance of a signal separation algorithm in white noise is presented. The special properties of nonlinear dynamical systems allow state estimation that improves exponentially with the number of observations but requires special processing techniques to achieve.>

The problem of modeling chaotic nonlinear dynamical systems using hidden Markov models is considered. A hidden Markov model for a class of chaotic systems is developed from noise-free observations of the output of that system. A combination of vector quantization and the Baum-Welch algorithm is used for training. The importance of this combined iterative approach is demonstrated. The model is then used for signal separation and signal detection problems. The difference between maximum likelihood signal estimation and maximum a posteriori signal estimation using a hidden Markov model is illustrated for a nonlinear dynamical system.>

This paper discusses the symbolic representation and manipulation of signals and signal processing operations in computers. Symbolic representation and manipulation of signals, in contrast to numerical representation and manipulation, is the manipulation of signal properties and signal descriptions, rather than the computation of signal values. We identify three classes of symbolic manipulations of signals - symbolic manipulation of signal properties, rearrangement of signal processing expressions, and manipulation of abstract signal classes - and we describe the Extended Signal Processing Language and Interactive Computing Environment (E-SPLICE), a system for the symbolic manipulation of signals. Examples of the symbolic manipulation of signals are given, including expression simplification, generation of equivalent forms of an expression, the manipulation of signal properties, the measurement of computational cost, and the selection of signal processing algorithm implementation.

This paper provides an interpretation of linear programming for spectral estimation [1,2] as building an MLM-like data-adaptive filter with separate weights on each pair of sensors rather than on each sensor individually. It is shown that this estimate becomes MLM when a positive semi-definiteness constraint is imposed on the matrix of weights and becomes linear programming when a non-negative response to sinusoids constraint is imposed on the matrix of weights. It is shown that, for some sensor array configurations, linear programming does indeed provide tighter bounds on the spectral power than MLM, linear programming can pick up lower level sources than MLM, and linear programming can reject more interfering sources than MLM.

This paper describes a system for speech analysis and enhancement which combines signal processing and symbolic processing in a closely coupled manner. The system takes as input both a noisy speech signal and a symbolic description of the speech signal. The system attempts to reconstruct the original speech waveform using symbolic processing to help model the signal and to guide reconstruction. The system uses various signal processing algorithms for parameter estimation and reconstruction.

Knowledge-based signal processing combines the symbolic manipulation facilities of artificial intelligence with the numerical methods of signal processing. This paper considers pitch detection as a problem for developing techniques for signal/symbol interaction. Specifically an overview of the design of the Pitch Detector's Assistant program is discussed, along with some strategies followed in the program. for the use of information derived from both signals and symbols.

In this paper we describe a system for connected word recognition in which a sentence in a formal language, uttered without pauses between words, is recognized by finding the grammatically well formed sequence of isolated word templates to which its distance is least. This is accomplished by means of a single monolithic algorithm in which temporal registration, segmentation and grammatical analysis are performed simultaneously. The algorithm is a syntax-directed version of the level building dynamic time warping algorithm of Myers and Rabiner. A test was conducted on a total of 208 sentences comprising 1781 words and spoken by two male and two female speakers. The sentences were composed from a 127 word vocabulary according to a moderately complex grammar and semantic structure appropriate to an airline information and reservation task. Test results revealed a 13% sentence error rate and a 6% word error rate.

The technique of dynamic time warping has proven itself reliable and robust for a wide variety of isolated word recognition tasks. Recently extensions of the algorithm have been investigated for application to the problem of connected word recognition. In this paper a level building technique is proposed for optimally aligning a test pattern, consisting of a sequence of connected words, with a sequence of isolated word reference patterns. This algorithm is shown to be significantly more efficient than the one proposed by Sakoe while solving the exact same problem. Implementation parameters for the level building algorithm are presented and the effectiveness of the proposed algorithm for connected digit recognition is experimentally verified.

Several variations on algorithms for dynamic time warping have been proposed for speech processing applications. In this paper two general algorithms that have been proposed for word spotting and connected word recognition are studied. These algorithms are called the fixed range method and the local minimum method. The characteristics and properties of these algorithms are discussed. It is shown that, in several simple performance evaluations, the local minimum method performed considerably better then the fixed range method. Explanations of this behavior are given and an optimized method of applying the local minimum algorithm to word spotting and connected word recognition is described.