quick navigator
Products
Technologies
Development Tools
Intel Recognition Primitives Library
*Features/Benefits
*System Requirements
*Performance Specifications
*What's New -- Revision History
*Demos
*Licensing
*Download and Install
*Technical Support
*Back to Intel Software Performance Products Home
Developer Home Contents Search Feedback Support Intel(r)

The Intel Recognition Primitives Library Features and Benefits

The Intel Recognition Primitives Library V3.1 provides developers of speech and character-recognition software a set of generic functions in the form of standard DLLs and static libraries. The generic functions are optimized for various Intel Architecture processors, such as Intel486, Pentium®, Pentium with MMX technology, Pentium Pro, and Pentium II processors. The recognition primitive functions address the most compute-intensive parts of typical speech- and character-recognition applications. The library includes functions for windowing, fast Fourier transforms, multi-layer perceptron, Hidden Markov Model-based functions (discrete, continuous, and semi-continuous), and many vector operations. The functions process 16-bit integer data, single precision and, floating-point data, as well as their complex variations. Because MMX technology accelerates integer arithmetic primarily, the floating-point functions do not use MMX technology.


Features Benefits
Highly optimized processor-specific DLLs and static libraries Optimal product performance and reduced time to market
C-Callable functions High-level language programming interface
Custom DLL builder utility Minimum memory footprint of DLL
Both integer and floating point functions Support for a wide range of needs
Processor detection and processor-specific DLL loading Single scalable executable

Content

Vector Operations Initialization, Arithmetic, Logical
Signal Processing Windowing, FFT
Speech Signal Processing Signal Pre-emphasis, Cepstral Analysis
Recognition Basics Dot Products, Euclidean Distances, Mahalanobis Distances, Gaussian Mixtures, MLP Neural Network, Vector Quantization/Kohonen Network.
Elastic Matching Dynamic Time Warping, Hidden Markov Models (Discrete, Continuous, and Semi-Continuous functions), Viterbi Algorithm.
Image Processing Pixel Arithmetic and Logical Operations, Image Geometric Transformations, Mask Convolution

* Legal Information © 1998 Intel Corporation