Go to the first, previous, next, last section, table of contents.


Statistical Routines

The speechtools package provies some statistical routines of the type that are frequently used in speech applications.

Simple statistics

These simple statistical classes offer some basic functionality

EST_Discrete

This class is for mapping a list of names to a set of integers. It was written for use in probability density functions which are used in the N-gram classes.

It may be initialised with a string of names (EST_StrList) and offers conversion between string names and the integer indexes and vice versa.

The class definition is in `include/EST_sinplestats.h'.

EST_SuffStats

This class allows the cummulation of sample numbers so various common simple statistics may be derived. It basically keeps the number of samples, sum and sum square of the numbers it accumulates so mean, variance and standard deviation man be calculated.


   EST_SuffStats a;

   for (i=0; i < list_size; i++)
      a+=list[i];

   cout << "n " << a.samples() << endl;
   cout << "mean " << a.mean() << endl;
   cout << "variance " << a.variance() << endl;
   cout << "stddev " << a.stddev() << endl;

EST_DiscreteProbDistribution

This class allows the cumulation of discretes keeping a count of their occurences. It may be initialised from a vocabulary in an EST_StrList or a EST_Discrete.

This class can accumulate samples (strings or ints). It keeps track of counts so that frequecies, probabilities, entropy can be easily calculated through the provided member functions.

The class is declared in `include/EST_simplestats.h'.

Multivariate statistics

to be written

Clustering

to be written


Go to the first, previous, next, last section, table of contents.