Intel® Integrated Performance Primitive Functions for Signal Processing

  • Vector initialization functions include functions that initialize the values of the elements of a vector. All vector’s elements can be initialized to zero or to another specified value. They can also be initialized to the value of a second vector. Sample-generating functions, filed in this same group, generate the tone samples, triangle samples, pseudo-random samples with uniform distribution, and pseudo-random samples with Gaussian distribution, and specific Jaehne test samples.

  • Essential vector functions perform logical and shift, arithmetic, conversion, companding, windowing, and statistical operations.

  • Filtering functions perform convolution and correlation operations, as well as linear and non-linear filtering.

  • Fourier transform functions perform the discrete Fourier transform (DFT), the fast Fourier transform (FFT), and the discrete cosine transform (DCT) of signal samples. This group also includes variations of the basic functions to support different application requirements.

  • Speech recognition functions perform a series of operations essential for speech recognition application. This group includes functions for generic arithmetic operations on vectors and matrices; feature processing functions that pre-process raw speech signals; functions for acoustic and language model evaluation, estimation, and adaptation; and functions for vector quantization and codebook operations commonly used in acoustic and language model compressions.

  • Audio coding functions include general purpose functions applicable in several codecs and a number of specific functions for MPEG-4 audio encoder and decoder.

  • MP3 audio decoding functions can be used to construct audio decoders compliant with the layer III portions of the ISO/IEC 11172-3 MPEG-1 and ISO/IEC 13818-3 MPEG-2 audio specifications. These international standards for perceptual coding (compression) of digital audio are most often denoted by the “MP3” acronym. The MP3 algorithm delivers high quality audio with bit rates as low as one-tenth of the original, making it a popular choice for cost-sensitive and bandwidth-constrained transmission and/or storage applications. At the same time, the MP3 decoder is characterized by a manageable computational complexity. As a result, MP3 has become the de facto standard audio compression technology for emergent portable and hand-held digital media, as well as for distribution of high-fidelity compressed audio over networks such as the Internet.

  • Fixed-accuracy transcendental mathematical functions of vector arguments take an input vector as argument, compute values of the respective elementary function element-wise, and return the results in an output vector.