jamie@1: /* libxtract feature extraction library
jamie@1: *
jamie@1: * Copyright (C) 2006 Jamie Bullock
jamie@1: *
jamie@1: * This program is free software; you can redistribute it and/or modify
jamie@1: * it under the terms of the GNU General Public License as published by
jamie@1: * the Free Software Foundation; either version 2 of the License, or
jamie@1: * (at your option) any later version.
jamie@1: *
jamie@1: * This program is distributed in the hope that it will be useful,
jamie@1: * but WITHOUT ANY WARRANTY; without even the implied warranty of
jamie@1: * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
jamie@1: * GNU General Public License for more details.
jamie@1: *
jamie@1: * You should have received a copy of the GNU General Public License
jamie@1: * along with this program; if not, write to the Free Software
jamie@1: * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301,
jamie@1: * USA.
jamie@1: */
jamie@1:
jamie@20: /** \mainpage
jamie@20: *
jamie@20: * LibXtract is a simple, portable, lightweight library of audio feature extraction functions. The purpose of the library is to provide a relatively exhaustive set of feature extraction primatives that are designed to be 'cascaded' to create a extraction hierarchies.
jamie@20: * For example, 'variance', 'average deviation', 'skewness' and 'kurtosis', all require the 'mean' of the input vector to be precomputed. However, rather than compute the 'mean' 'inside' each function, it is expected that the 'mean' will be passed in as an argument. This means that if the user wishes to use all of these features, the mean is calculated only once, and then passed to any functions that require it.
jamie@20: *
jamie@20: * This philosophy of 'cascading' features is followed throughout the library, for example with features that operate on the magnitude spectrum of a signal vector (e.g. 'irregularity'), the magnitude spectrum is not calculated 'inside' the respective function, instead, a pointer to the first element in an array containing the magnitude spectrum is passed in as an argument.
jamie@20: *
jamie@20: * Hopefully this not only makes the library more efficient when computing large numbers of features, but also makes it more flexible because extraction functions can be combined arbitrarily (one can take the irregularility of the Mel Frequency Cepstral Coefficients for example).
jamie@20: *
jamie@20: */
jamie@20:
jamie@1: #ifndef XTRACT_H
jamie@1: #define XTRACT_H
jamie@1:
jamie@1: #ifdef __cplusplus
jamie@1: extern "C" {
jamie@1: #endif
jamie@1:
jamie@21: /**
jamie@21: * \file libxtract.h
jamie@21: * \brief main header file and API definition
jamie@1: */
jamie@1:
jamie@1: #include "xtract_scalar.h"
jamie@1: #include "xtract_vector.h"
jamie@1: #include "xtract_delta.h"
jamie@1: #include "xtract_types.h"
jamie@1: #include "xtract_macros.h"
jamie@1:
jamie@20: /** \defgroup libxtract API
jamie@20: *
jamie@20: * Defines a very simple API that provides access to the functions in the library
jamie@20: * @{
jamie@20: */
jamie@20:
jamie@43: #define XTRACT_FEATURES 45
jamie@30:
jamie@1: #define LOG_LIMIT 10e-10
jamie@5: #define VERY_BIG_NUMBER 2e10
jamie@1: #define SR_LIMIT 192000
jamie@1: #define BARK_BANDS 26
jamie@1:
jamie@1: /** \brief Enumeration of features, elements are used as indixes to an array of pointers to feature extracton functions */
jamie@1: enum features_ {
jamie@1: MEAN,
jamie@1: VARIANCE,
jamie@1: STANDARD_DEVIATION,
jamie@1: AVERAGE_DEVIATION,
jamie@1: SKEWNESS,
jamie@1: KURTOSIS,
jamie@29: CENTROID,
jamie@1: IRREGULARITY_K,
jamie@1: IRREGULARITY_J,
jamie@1: TRISTIMULUS_1,
jamie@1: TRISTIMULUS_2,
jamie@1: TRISTIMULUS_3,
jamie@1: SMOOTHNESS,
jamie@1: SPREAD,
jamie@1: ZCR,
jamie@1: ROLLOFF,
jamie@1: LOUDNESS,
jamie@1: FLATNESS,
jamie@1: TONALITY,
jamie@1: CREST,
jamie@1: NOISINESS,
jamie@1: RMS_AMPLITUDE,
jamie@1: INHARMONICITY,
jamie@1: POWER,
jamie@1: ODD_EVEN_RATIO,
jamie@1: SHARPNESS,
jamie@1: SLOPE,
jamie@43: LOWEST,
jamie@1: HPS,
jamie@9: F0,
jamie@43: FAILSAFE_F0,
jamie@1: FLUX,
jamie@1: ATTACK_TIME,
jamie@1: DECAY_TIME,
jamie@30: DELTA_FEATURE,
jamie@30: AUTOCORRELATION,
jamie@30: AMDF,
jamie@30: ASDF,
jamie@30: BARK_COEFFICIENTS,
jamie@30: PEAKS,
jamie@30: MAGNITUDE_SPECTRUM,
jamie@30: AUTOCORRELATION_FFT,
jamie@30: MFCC,
jamie@38: DCT,
jamie@38: HARMONICS
jamie@1: };
jamie@1:
jamie@1: /** \brief Enumeration of feature types */
jamie@1: enum feature_types_ {
jamie@1: SCALAR,
jamie@1: VECTOR,
jamie@1: DELTA
jamie@1: };
jamie@1:
jamie@1: /** \brief Enumeration of mfcc types */
jamie@1: enum mfcc_types_ {
jamie@1: EQUAL_GAIN,
jamie@1: EQUAL_AREA
jamie@1: };
jamie@1:
jamie@1: /** \brief Enumeration of return codes */
jamie@1: enum return_codes_ {
jamie@1: SUCCESS,
jamie@1: MALLOC_FAILED,
jamie@1: BAD_ARGV,
jamie@12: BAD_VECTOR_SIZE,
jamie@38: NO_RESULT,
jamie@38: FEATURE_NOT_IMPLEMENTED
jamie@1: };
jamie@1:
jamie@1: /**
jamie@1: *
jamie@2: * \brief An array of pointers to functions that perform the extraction
jamie@1: *
jamie@2: * \param *data: a pointer to the start of the input data (usually the first element in an array)
jamie@1: *
jamie@2: * \param N: the number of elements to be processed
jamie@1: *
jamie@2: * \param *argv: an abitrary number of additional arguments, used to pass additional parameters to the function being called
jamie@1: *
jamie@2: * \param *result: a pointer to the first element in the result
jamie@1: *
jamie@1: * Each function will iterate over N array elements, the first of which is
jamie@2: * pointed to by *data. It is up to the calling function to ensure that the array is in the format expected by the function being called.
jamie@1: *
jamie@1: * For scalar and delta features, *result will point to a single value.
jamie@1: *
jamie@1: * For vector features it will point to the first element in an array.
jamie@1: *
jamie@1: * Memory for this array must be allocated and freed by the calling
jamie@1: * function.
jamie@1: *
jamie@21: * All functions return an integer error code as descibed in the enumeration
jamie@21: * return_codes_
jamie@2: *
jamie@21: * The preprocessor macro: XTRACT must be defined before this can be used
jamie@9: *
jamie@2: * example:
jamie@21: * \verbatim
jamie@21: #include
jamie@21: #define XTRACT
jamie@21: #include "libxtract.h"
jamie@21:
jamie@21: main () {
jamie@21: float values[] = {1.0, 2.0, 3.0, 4.0, 5.0};
jamie@21: int N = 5;
jamie@21: float mean;
jamie@21:
jamie@21: xtract[MEAN]((void *)values, N, NULL, &mean);
jamie@21:
jamie@21: printf("Mean = %.2f\n", mean);
jamie@21: }
jamie@21: \endverbatim
jamie@9: * The calling function may additionally make some tests against the value returned by xtract
jamie@9: *
jamie@2: */
jamie@9: #ifdef XTRACT
jamie@43: extern int(*xtract[XTRACT_FEATURES])(const float *data, const int N, const void *argv, float *result);
jamie@1:
jamie@26: /** \brief An array of pointers to function help strings
jamie@26: *
jamie@26: * Defined in libxtract.c. As a minimum this will contain pointers to the names of all of the feature extraction functions in the library. This is intended as a 'quick reference' to be queried as necessary.
jamie@26: */
jamie@36: extern char *xtract_help_strings[XTRACT_FEATURES];
jamie@38:
jamie@38: /** \brief An array of pointers to strings giving function output units
jamie@38: *
jamie@38: * Defined in libxtract.c. This contains pointers to strings giving the output units of all of the feature extraction functions in the library. This is intended as a 'quick reference' to be queried as necessary.
jamie@38: */
jamie@38: extern char *xtract_units[XTRACT_FEATURES];
jamie@38:
jamie@28: #endif
jamie@26:
jamie@2: /** \brief A structure to store a set of n_filters Mel filters */
jamie@1: typedef struct xtract_mel_filter_ {
jamie@1: int n_filters;
jamie@1: float **filters;
jamie@1: } xtract_mel_filter;
jamie@1:
jamie@2: /** \brief A function to initialise a mel filter bank
jamie@2: *
jamie@2: * It is up to the caller to pass in a pointer to memory allocated for freq_bands arrays of length N. This function populates these arrays with magnitude coefficients representing the mel filterbank on a linear scale
jamie@2: */
jamie@1: int xtract_init_mfcc(int N, float nyquist, int style, float freq_max, float freq_min, int freq_bands, float **fft_tables);
jamie@1:
jamie@2: /** \brief A function to initialise bark filter bounds
jamie@2: *
jamie@2: * A pointer to an array of BARK_BANDS ints most be passed in, and is populated with BARK_BANDS fft bin numbers representing the limits of each band
jamie@2: */
jamie@1: int xtract_init_bark(int N, float nyquist, int *band_limits);
jamie@1:
jamie@1:
jamie@1: /* Free functions */
jamie@1:
jamie@20: /** @} */
jamie@20:
jamie@1: #ifdef __cplusplus
jamie@1: }
jamie@1: #endif
jamie@1:
jamie@1: #endif