Robert J. Sluyter

Measurement of pitch in speech: An implementation of Goldstein’s theory of pitch perception

Recent developments in hearing theory have resulted in the rather general acceptance of the idea that the perception of pitch of complex sounds is the result of the psychological pattern recognition process. The pitch is supposedly mediated by the fundamental of the harmonic spectrum which fits the spectrum of the complex sound optimally. The problem of finding the pitch is then equivalent to finding the best harmonic match. Goldstein [J. Acoust. Soc. Am. 54, 1496-1516 (1973)] has described an objective procedure for finding the best fit for stimuli containing relatively few spectral components. He uses maximum likelihood criterion. Application of this procedure to various data on the pitch of complex sounds yielded good results. This motivated our efforts to apply the pattern recognition theory of pitch to the problem of measuring pitch in speech. Although we were able to follow the main line of Goldsteins procedure, some essential changes had to be made. The most important is that in our implementation not all spectral components of the complex sound have to be classified as belonging to the harmonic pattern. We introduced a harmonics sieve to determine whether components are rejected or accepted at a candidate pitch. A simple criterion, based on the components accepted and rejected, led to the decision on which candidate pitch was to be finally selected. The performance and reliability of this psychoacoustically based pitch meter were tested in a LPC-vocoder system.

Speech codec for the European mobile radio system

In 1991 a digital mobile radio system will be introduced in Europe. The speech codec to be used as the standard is presented. The coding scheme which has been selected by the CEPT Groupe-Speciale-Mobile (GSM) as a result of formal subjective listening tests, is based on the regular-pulse excitation LPC technique (RPE-LPC) combined with long-term prediction (LTP). The so-called RPE-LTP codec has a net bit rate of 13 kbit/s. The algorithm and the experimental implementations based on different VLSI signal processors are described and demonstrated.<<ETX>>

A programmable 1400 MOPS video signal processor

A description is given of a digital, general-purpose, programmable IC especially designed for flexible processing of video signals. The IC is implemented in a 1.6 μm CMOS technology and operates at 27 MHz while performing over 1400 MOPS (million user-programmable operations per second) at an instruction rate of 270 MIPS. A novel design method is described which blends silicon compiler concepts with conventional full-custom design techniques

A novel method for pitch extraction from speech and a hardware model applicable to vocoder systems

In this paper we describe a new and reliable concept for pitch extraction from speech which has been implemented in hardware and applied in a channel vocoder. Rather than looking at pitch extraction as a purely technical problem, the considered new technique is based on a recently developed model of pitch perception. The main functional ingredients of the concept are spectral analysis,the extraction of component frequencies from the spectrum and subsequent harmonic pattern recognition. In terms of hardware the frequency analysis is performed by a DFT processor, while the component extraction and harmonic pattern recognition are carried out by a microprocessor configuration.

A regular-pulse excited linear predictive codec

Abstract A 16 kbit/s speech codec with low complexity and low signal delay is presented which is a special version of the Regular-Pulse Excitation LPC approach (RPE-LPC). This proposal is the basis for the codec standard which will be used in the future Pan-European digital mobile telephone system. An experimental hardware model is described.

A general-purpose video signal processor: architecture and programming

Programming aspects of a new digital, flexible processor especially designed for the effective processing of real-time video signals are addressed. The modular architecture contains a number of programmable, pipelined processing elements. A programmable crossbar switch provides a flexible interconnection between the processing elements. The programs can be constructed with the aid of graphical, interactive tools that abstract from hardware details. Efficient mapping algorithms that automate parts of the programming trajectory have been designed. Data are presented on the hardware utilization that was achieved by applying the tools to a number of video algorithms.<<ETX>>

A programmable video signal processor

A description is given of a digital, general-purpose, programmable processor chip, especially designed for effective processing of video signals. The architecture is described in terms of three hierarchical levels: the system level, describing how video signal processing tasks are realized on the basis of only one type of processor chip; the chip level, describing the modular architecture with the number of modules on the chip to be tuned to the available technology; and the module level, describing the various processing elements, how they communicate, and how they are controlled by so-called cyclo-static programs. For program development, software support tools have been designed, in a top-down mapping trajectory, that start at the signal-flow-graph level and result in the program code for the processor(s).<<ETX>>

DFT-Vocoder using harmonic-sieve pitch extraction

The DFT-vocoder is based on speech analysis and synthesis using the discrete Fourier transform (DFT). Analysis is done using hopping-DFT and spectral parameters are obtained by a piece-wise constant approximation of the amplitude spectrum. The harmonic-sieve technique for pitch extraction combines very well with this scheme because it is based on hopping-DFT as well. Synthesis is achieved by convolution of the generated excitation signal with the inverse-DFT of the reconstructed piece-wise constant amplitude spectrum. In this paper a 2400 bit/s implementation of the DFT-vocoder is discussed.

Improvements of the harmonic-sieve pitch extraction scheme and an appropriate method for voiced-unvoiced detection

Continued investigations on the harmonic-sieve pitch extractor have led to several major improvements, such as e.g. even less frequent occurrence of pitch errors, especially for speech with low pitch, and extreme robustness with respect to low-frequency cut-off in the input speech. Additionally, a voiced-unvoiced detector is presented which is based on tracking of spectral intensity variations of subsequent speech segments. It can be incorporated in the pitch extractor with very little overhead.