gauss_example_32f_select.txt

Intel开发的IPP库的应用实例

                                       to delta calculation
                                       number of calculated probabilities is returned
                                       or -1 for wrong arguments
int State_Number(void)               - get state number in calculator (state_number)
int Max_Gauss(void)                  - get maximal Gaussians in mixture (max_gauss)
int Feature_Length(void)             - get feature vector length in calculator
float *Mean(int state)               - get pointer to first element of first mean of state
float *Var(int state)                - get pointer to first element of first var of state
float *Det(int state)                - get pointer to first det of state
float *Weight(int state)             - get pointer to first weight of state
void  Get_Statistics(Len_Stat *stat) - get statistics of whole number of gaussian component
                                       and number of calculated components. It's necessary
                                       for case with LogGauss Selection. In other cases it
                                       returns 0.
int   Init_Select_Gauss(char *file, int cdbksize, int num_clust, float val,
           int num_vec, bool is_row) - internal structures initialization for Gaussian Selection
                                       algorithm. Does nothing if the calculation hint is not equal
                                       to calcVecS. This function initializes Select_Gauss object
                                       that contains codebook and global shortlist and is used for
                                       local shortlist calculation. If codebook with required
                                       number of cluster is not in parameter file it is built
                                       from Gaussian means and grand variance and is written to
                                       HMM parameter file. Other arguments determine Gaussain
                                       Selection algorithm features.


        Feat_Calc class

Calculates feature vectors for probability calculator. Fake producer provides zero
feature vectors. File producer takes data from HTK feature file with float data.
Two modes of feature processing are supported: raw features without energy (as for
the first decoder pass) and CMS&VN'ed features with energy (as for the second pass).
Delta or double delta are calculated if necessary.

Feat_Calc (int space_dim, int max_delay) - constructor
                                       Each feature vector of length space_dim is
                                       appended to the length (space_dim+3)&(~3)
                                       for 16 byte alignment
                                       max_delay is used to imitate low delay
                                       feature producer - features are calculated by
                                       portions of length no more than max_delay.
                                       If deltas are calculated delta_delay =
                                       delta_window*(delta_number+1) is added.
                                       delta_delay is 0 for fake feature producer
                                       If max_delay<=0 features for all utterance
                                       could be calculated at once
                                       Feature producer has no internal buffer and
                                       calculates features in consumer's buffer
~Init_Feat(void)                     - destructor
int Init_Feat(int frame_number)      - internal structures initialization for fake
int Init_Feat(char *file)              and SDT based feature producers
                                       Fake feature producer provides an utterance
                                       of frame_number frames, each frame is zero
                                       vector of length space_dim
                                       HTK file based feature producer reads raw feature
                                       vectors of length raw_dim from HTK feature file
                                       Feature vector elements assumed to be floats.
                                       space_dim should be equal to raw_dim, raw_dim*2,
                                       raw_dim*3, (raw_dim-1), (raw_dim-1)*2 or
                                       (raw_dim-1)*3. If necessary last element of
                                       raw feature is deleted and deltas or double
                                       deltas with delta_window=2 are calculated.
                                       >= 0 if OK, otherwise failed

After successful constructor and Init_Feat calls producer is ready

bool Ready(void)                     - true if producer is ready, false otherwise
int  Get_Feature(float *dst, int len, bool* end)
                                     - puts calculated featurs to consumers buffer dst
                                       len features is requested, the number of
                                       fully calculated features is returned
                                       if utterance is not ended  they are followed by
                                       delta_delay partially calculated feature vectors
                                       The first partially calculated feature vector is
                                       passed as dst for the next feature portion
                                       if the calculated portion ends the utterance
                                       end[0] is true otherwise false
int Feature_Length(void)             - get output feature vector length
int Delta_Delay(void)                - get delta_delay of feature producer
                                       Probability calculator requests this number to
                                       provide correct interaction with probability
int Write_LogHeader(FILE *flog)      - writes HTK feature file header to log file
                                       feature vector length includes deltas or
                                       double deltas if any
                                       calculator


        Select_Gauss class

Contains codebook and shortlist that are used for LogGauss Selection. Also contain
function to manage indexes (for input vectors) and Sign table (it's necessary for
function ippsLogGaussMixture_Select_32f_D2 in probability calculation)

Select_Gauss (int cdbk_size, bool is_row, int n_vec, float val, int num_clust, int spc_dim)
                                     - constructor, simply fill zero in pointers
                                       and set values to zero
                        cdbk_size    - number of clusters in codebook,
                        is_row       - defines method of global shortlist filling
                                       if true - shortlist is filled by rows (one or more
                                       components of each mixture are chosen to be active
                                       for each codebook cluster)
                                       if false - shortlist is filled by columns (each
                                       component of each mixture are chosen to be active
                                       for one or more codebook clusters)
                        n_vec        - number of active components if is_row is true or
                                       number of clusters for which component is active
                                       if is_row is false
                                       if num_vec is 0, variable number of components are
                                       activated according to threshold value val.
                                       used while shortlist building
                        val          - defines component activating threshold
                                       if is_row is true then for each cluster
                                       components with distance to centroid less then
                                       minimal*val are activated
                                       if is_row is false then each component is activated
                                       for clusters with distance to centroid less then
                                       minimal*val
                                       used while shortlist building
                        num_clust    - mixture components belonging to num_clust clusters
                                       closest to input feature vector are calculated
                                       used while mixture calculating
                        spc_dim      - feature space dimension
~Select_Gauss(void)                  - destructor
int Init_SelectGauss_Simple(int m_len, int m_gauss, int gss_num, foat *tmpcodebook, float *weights,
                            float *meanSet, Mixture *states, int states_num)
                                     - initializes internal structure such as codebook using
                                       input arrays, which have been read from param files by
                                       function ReadAndInitCdbk.
                        m_len        - maximum length of input vectors to quantize
                        m_gauss      - maximum number of gaussian mixture components
                        gss_num      - number of gaussians in the model
                        tmpcodebook  - array to store created codebook
                        weights      - array of weights for weighted Euclidean distance
                        meanSet      - array of means values, it's source for creating codebook
                        states       - array of components for each state
                        states_num   - number of states

int Init_SelectGauss_Full(int m_len, int m_gauss, int gss_num, foat *tmpcodebook, float *weights,
                            float *meanSet, Mixture *states, int states_num)
                                     - initializes internal structure and create codebook using
                                       input arrays, which have been read from param files by
                                       function CreateInitSaveCdbk. Also return created codebook
                                       vectors and weights vector to parent function to save it
                                       into the param file.
                        m_len        - maximum length of input vectors to quantize
                        m_gauss      - maximum number of gaussian mixture components
                        gss_num      - number of gaussians in the model
                        tmpcodebook  - array to store created codebook
                        weights      - array to store weights for weighted Euclidean distance
                        meanSet      - array of means values, it's source for creating codebook
                        states       - array of components for each state
                        states_num   - number of states

int CreateShortList(float* means)    - fills shortlist using means array and created by previous
                                       functions codebook
int CommonInit(float* means)         - mallocs arrays for select_gauss object and call previous
                                       function to fill shortlist
int WriteVQ(float* feature, int shift, int len)
                                     - quantize input vector according to the shortlist and input
                                       parameters
                        feature      - input vector
                        shift        - shift in the indexes vector
                        len          - len of input vectors to quantize

int    ShiftIndxBegin(int len)       - shifts indexes in the index vector. It's necessary to
                                       accord the StepForward function when we shift and reread
                                       input vector

int FillSign(int esnmix, int shiftbits, int num,int shift)
                                     - fills Sign table using vector of indexes, shortlist and
                                       it's own parameters.
                        esnmix       - number of mixture components
                        shiftbits    - shift in shortlist table in bits
                        num          - number of indexes to fill
                        shift        - shift in vector of indexes