Abstracts - faqs.org

Abstracts

Electronics

Search abstracts:
Abstracts » Electronics

Neural network models of sensory integration for improved vowel recognition

Article Abstract:

Multiple sources of speech information can be used by neural networks to improve speech recognition in the presence of noise. Humans compensate for loss of acoustical information by using information from mouth movements, facial expressions and gestures. Massively parallel artificial neural networks can combine information from several sources to improve speech recognition. Feedforward and recurrent networks are trained to estimate acoustics associated with a visual speech signal. Once data from the sources is combined, the effective signal to noise ratio is increased and recognition is more accurate. Networks can also be used to directly determine vowels from visual signals. The neural nets perform well compared to other pattern recognition techniques and human performance.

Author: Yuhas, Ben P., Goldstein, Moise H., Jr., Sejnowski, Terrence J., Jenkins, Robert E.
Publisher: Institute of Electrical and Electronics Engineers, Inc.
Publication Name: Proceedings of the IEEE
Subject: Electronics
ISSN: 0018-9219
Year: 1990
Neural networks, Speech recognition, Voice Recognition, technical, Signal to Noise Ratio, Neural Network, Pattern Matching, Visualization

User Contributions:

Comment about this article or add new information about this topic:

CAPTCHA


Switched-Capacitor Circuit Design

Article Abstract:

High quality analog filters are realized as passive LCR circuits. Inductors are large and noisy and hence unsuitable for miniaturization. Fully integrated filters are being manufactured using MOS technology. Switched-capacitor integrators are the ones used mostly in many applications. These integrators are the basic blocks of filters. Several design techniques for switched-capacitors are discussed. A block diagram shows a tracking filter realization.

Author: Gregorian, R., Martin, K.W., Temes, G.C.
Publisher: Institute of Electrical and Electronics Engineers, Inc.
Publication Name: Proceedings of the IEEE
Subject: Electronics
ISSN: 0018-9219
Year: 1983
Integrated circuits, Circuit design, Design, Capacitors, Analog to Digital Converter, Circuitry, Switching, Filters, Filtering, Programmable, Arrays, Oscillators

User Contributions:

Comment about this article or add new information about this topic:

CAPTCHA


An Improved Conference Circuit

Article Abstract:

The overall performance of a conferencing circuit is improved by minimizing return loss. Alternating inverting amplifier stages at the summing junctions of a signal addition conference circuit cancels out the effect of feedback voltages. Signal addition conferencing circuits are illustrated.

Author: Pitke, M.V., Chandrasekaren, T.
Publisher: Institute of Electrical and Electronics Engineers, Inc.
Publication Name: Proceedings of the IEEE
Subject: Electronics
ISSN: 0018-9219
Year: 1983
Teleconferencing, Performance, Signal Propagation

User Contributions:

Comment about this article or add new information about this topic:

CAPTCHA


Subjects list: Communications Circuits
Similar abstracts:
  • Abstracts: A tutorial on hidden Markov models and selected applications in speech recognition. A Tutorial on Computer-Aided Analysis of Switched-Capacitor Circuits
  • Abstracts: Internetwork Protocol. Status of CCITT Description Techniques and Application to Protocol Specification. Fundamentals of the Layer Service Definitions and Protocol Specifications
  • Abstracts: Layer assignment for printed circuit boards and integrated circuits. Software maintenance: the need for standardization
  • Abstracts: Radio determination satellite service. Customer cost of electric service interruptions
  • Abstracts: Data General Desktop Generation Model 10: Architecture and Implementation. Business Data Usage of OSI
This website is not affiliated with document authors or copyright owners. This page is provided for informational purposes only. Unintentional errors are possible.
Some parts © 2025 Advameg, Inc.