A classifier to make the voiced/unvoiced (V/UV) decision in speech analysis which performs with an error rate of less than half of a percent is presented. The decision making process is viewed as a pattern recognition problem in which a number of features can be used to make the classification. Training is accomplished using a nonparametric, nonstatistical technique. In order to obtain a classifier which would make the correct decision for a variety of speakers and to determine which of the features under consideration should be used, a procedure for interleaving the contributions of the feature and speaker sets was developed. This procedure is presented in terms of the notions of covering and satisfaction. The failure of a classifier to cover a set of speakers indicates that more training information from those speakers is necessary to define the classifier. The failure of the classifier to satisfy a set of speakers indicates that the performance of the classifier could be improved by the use of more features in making the V/UV decision. In the training procedure, covering and satisfaction were attained on successively larger sets of speakers, with the result that a classifier was obtained which could correctly make the V/UV decision for all of the speakers used in testing, including those not used in the training process.

A procedure for using pattern classification techniques to obtain a voiced/Unvoiced classifier

The system described in this paper is subdivided into three main steps: pitch extraction, segmentation, and formant analysis. The pitch extractor uses an adaptive digital filter in time-domain transforming the speech signal into a signal similar to the glottal waveform. Using the levels of the speech signal and the differenced signal as parameters in time domain, the subsequent segmentation algorithm derives a signal parameter which describes the speed of articulatory movement. From this, the signal is divided into "stationary" and "'transitional" segments; one stationary segment is associated to one phoneme. For the formant tracking procedure, a subset of the pitch periods is selected by the segmentation algorithm and is transformed into frequency domain. The formant tracking algorithm uses a maximum detection strategy and continuity criteria for adjacent spectra. After this step, the total parameter set is offered to an adaptive universal pattern classifier which is trained by selected material before working. For stationary phonemes, the recognition rate is about 85 percent when training material and test material are uttered by the same speaker. The recognition rate is increased to about 90 percent when segmentation results are used.

https://ieeexplore.ieee.org/iel6/29/26116/01162771.pdf

A pitch-synchronous digital feature extraction system for phonemic recognition of speech

The existence is reported of large reversible changes in the excess current of germanium, silicon, and gallium arsenide tunnel diodes subjected to highly localized stress. The relationship of this effect to the anisotropic stress effect recently reported by Rindner and Braun in conventional p‐n junctions is discussed. It is suggested that both effects may arise from deep‐lying electronic states associated with strain‐induced lattice defects in the junction region. A consistent interpretation is presented of the tunnel diode excess current in terms of strain‐induced intermediate tunneling states, and of the conventional diode current in terms of strain‐induced generation‐recombination centers.

Anisotropic Stress Effect on the Excess Current in Tunnel Diodes

A speech processing system has been developed which is capable of providing an accurate indication of whether or not a given speech segment is voiced or unvoiced. In comparison to other existing techniques, this one avails itself for easy implementation, and it offers a realiable voiced/unvoiced (V/UV) decision even in the presence of some competing speech and noise sources. In addition, the V/UV decision is achieved in real time with time delays of \leq 4 ms going from voiced to unvoiced speech, and \leq 2 ms going from unvoiced to voiced speech.

Reliable voiced/Unvoiced decision

Monaural speech separation using GA-DNN integration scheme

The processing of speech involves the analysis, coding, decoding, and synthesis of speech sounds. The speech analyzer consists of normalizers, syllable and syblet segmenters, sound recognizers, sequencers, adapters, and memories which convert the speech elements into a code. The speech synthesizer converts the code to speech by reproducing prerecorded speech elements. There are many applications for the speech analyzer and synthesizer ranging from limited vocabulary to complete communication systems.

Speech processing techniques and applications

A tunnel‐diode transducer element is described in which stress is applied normal to the diode junction by a small probe. Static measurements on these elements have yielded strain gauge factors as high as 7300. A number of experimental microphones have been built to evaluate the transducers. Response frequency characteristics and S/N ratios have been determined for several of these microphones. The most sensitive microphone had a sensitivity of −31 dBm and a S/N of 45 dB at resonance. These microphones have been operated both as variable resistance transducers and as modulated negative resistance oscillator transducers. A description and analysis is given of these two basic circuit configurations. In their present form tunnel‐diode transducers are fragile; further effort will be required to produce a rugged, practical transducer element.

Experimental Tunnel‐Diode Electromechanical Transducer Elements and Their Use in Tunnel‐Diode Microphones

A speech analyzer and synthesizer with a capacity of 200 speech elements has been developed with the objective of evaluating the performance. The speech analyzer consisting of normalizers, syllable and syblet segmenters, sound recognizers, sequencers, and memories converts the speech elements into a code. The speech synthesizer converts the code to speech by reproducing prerecorded speech elements. The advantages of the voice‐communication system employing the analyzer and synthesizer are the high intelligibility and the low transmission rate of only 23 bit/sec.

Edward S. Rogers

Papers

Speech processing techniques and applications

Experimental Tunnel‐Diode Electromechanical Transducer Elements and Their Use in Tunnel‐Diode Microphones

Research Towards a High Efficiency Voice Communication System

Research towards a High‐Efficiency Voice‐Communication System