Platt's binary SVM Probablistic Output: an improvement from Lin et al.

Platt's binary SVM Probablistic Output: an improvement from Lin et al.

Hermansky 1990, JASA , scale corrected to max == 1 */

Implementation of an inverse lattice filter, processes a single value per call */

Implementation of a lattice filter, processes a single value per call */

sinc function (modified) : (sin 2pi*x) / x */

convert lp coefficients to cepstra HTKbook , equation 5.11 */

computes formant frequencies and bandwidths from lpc polynomial roots return value: number of valid formants detected from given lpc polynomial r

peak enhancement in a linear magnitude spectrum */

smooth a magnitude spectrum (linear) */

Bartlett-Hann window */

Blackman-Harris window */

Blackman window */

gaussian window ...??? */

compute A/B , store in C */

fix roots to inside the unit circle */ ensure all roots are within the unit circle if abs(root) > 1 (outside circle) then root = 1 / root roots is a

median of vector x (workspace can be a pointer to an array of 2*N (!) FLOAT_DMEMs which is used to sort the data in x without changing x) (if work

compute L1 norm (absolute sum) of given vector x */

void smileRnn_readJsonFile(rapidjson::Document *doc, const std::string &filename);

Estimate the probability of a value x belonging to a given pmf h (the pmf must be normalised to sum 1!) */

QuickSort algorithm for a FLOAT_DMEM array with nEl elements */

buffer pointer will be updated to match current position caller must have a copy in order to properly free it

construct and solve various formulations

construct and solve various formulations

Result was a success

Cross-validation decision values for probability estimates

Cross-validation decision values for probability estimates

Stratified cross validation

new function **/

Return parameter of a Laplace distribution

Return parameter of a Laplace distribution

get keys of associative array fields

sets the read pointer to writePtr - N