Hierarchical Hidden Markov Model for Continuous Recognition and Regression (Multi-class) More...

#include <xmmHierarchicalHmm.hpp>

Inheritance diagram for xmm::HierarchicalHMM:

Collaboration diagram for xmm::HierarchicalHMM:

Public Member Functions
	HierarchicalHMM (bool bimodal=false)
	Constructor. More...

	HierarchicalHMM (HierarchicalHMM const &src)
	Copy Constructor. More...

	HierarchicalHMM (Json::Value const &root)
	Constructor from Json Structure. More...

HierarchicalHMM &	operator= (HierarchicalHMM const &src)
	Assignment. More...

Class Manipulation
virtual void	removeClass (std::string const &label)
	Remove a specific class by label. More...

virtual void	clear ()
	Remove All models. More...

Accessors
void	addExitPoint (int state, float proba)

Performance
virtual void	reset ()
	Resets the fitering process (recognition or regression) More...

virtual void	filter (std::vector< float > const &observation)
	filters a incoming observation (performs recognition or regression) More...

Json I/O
virtual Json::Value	toJson () const
	Write the object to a JSON Structure. More...

virtual void	fromJson (Json::Value const &root)
	Read the object from a JSON Structure. More...

Class Manipulation
unsigned int	size () const
	Get the number of classes in the model. More...

bool	hasClass (std::string const &label) const
	Checks if a class exists. More...

int	getIndex (std::string const &label) const
	Checks if a class exists. More...

Training
bool	trained () const
	Checks if the model is trained (training finished and not empty) More...

bool	training () const
	Checks if the model is still training. More...

virtual void	train (TrainingSet *trainingSet)
	Train all classes from the training set passed in argument. More...

virtual void	train (TrainingSet *trainingSet, std::string const &label)
	Train a specific class from the training set passed in argument. More...

void	cancelTraining ()
	Cancels the training of all models. More...

void	cancelTraining (std::string const &label)
	Cancels the training of a given class. More...

Python File I/O
void	writeFile (char *fileName) const
	write method for python wrapping ('write' keyword forbidden, name has to be different) More...

void	readFile (char *fileName)
	read method for python wrapping ('read' keyword forbidden, name has to be different) More...

std::string	__str__ () const
	"print" method for python => returns the results of write method More...

Static Public Member Functions
static const double	DEFAULT_EXITTRANSITION ()
	Default exit transition for the highest level. More...

Public Attributes
Results< HMM >	results
	Convert to bimodal HierarchicalHMM in place. More...

std::vector< double >	prior
	Prior probabilities of the models. More...

std::vector< double >	exit_transition
	exit probabilities of the model (probability to finish and go back to the root) More...

std::vector< std::vector< double > >	transition
	Transition probabilities between models. More...

std::shared_ptr< SharedParameters >	shared_parameters
	Set of Parameters shared among classes. More...

Configuration< HMM >	configuration
	Configuration (default and class-specific parameters) More...

EventGenerator< TrainingEvent >	training_events
	Generator for training process events. More...

std::map< std::string, SingleClassHMM >	models
	models stored in a map. Each Model is associated with a label More...

Protected Member Functions
virtual void	joinTraining ()
	Finishes the background training process by joining threads and deleting the models which training failed. More...

void	updateTransitionParameters ()
	update high-level parameters when a new primitive is learned More...

void	updatePrior ()
	ergodic learning update high-level prior probabilities -> equal prior probs More...

void	updateTransition ()
	ergodic learning: update high-level transition matrix More...

void	updateExitProbabilities ()
	Update exit probabilities of each sub-model. More...

virtual void	addModelForClass (std::string const &label)
	Update training set for a specific label. More...

void	normalizeTransitions ()
	Normalize segment level prior and transition matrices. More...

void	forward_init (std::vector< float > const &observation)
	Initialization of the Forward Algorithm for the hierarchical HMM. see: Jules Francoise. Realtime Segmentation and Recognition of Gestures using Hierarchical Markov Models. Master’s Thesis, Université Pierre et Marie Curie, Ircam, 2011. [http://articles.ircam.fr/textes/Francoise11a/index.pdf]. More...

void	forward_update (std::vector< float > const &observation)
	Update of the Forward Algorithm for the hierarchical HMM. see: Jules Francoise. Realtime Segmentation and Recognition of Gestures using Hierarchical Markov Models. Master’s Thesis, Université Pierre et Marie Curie, Ircam, 2011. [http://articles.ircam.fr/textes/Francoise11a/index.pdf]. More...

void	likelihoodAlpha (int exitNum, std::vector< double > &likelihoodVector) const
	get instantaneous likelihood More...

void	updateResults ()
	Update the results (Likelihoods) More...

void	onTrainingEvent (TrainingEvent const &e)
	Monitors the training of each Model of the group. More...

void	checkTraining () const
	Checks if the Model is still training. More...

void	checkConfigurationChanges () const
	Look for configuration changes and throws an exception if the model is not up to date. More...

Protected Attributes
bool	forward_initialized_
	Defines if the forward algorithm has been initialized. More...

std::vector< double >	frontier_v1_
	intermediate Forward variable (used in Frontier algorithm) More...

std::vector< double >	frontier_v2_
	intermediate Forward variable (used in Frontier algorithm) More...

std::map< std::string, std::thread >	training_threads_
	Training Threads. More...

std::atomic< bool >	cancel_required_
	locks the Model while the models are training Used in cancel method More...

std::atomic< bool >	is_training_
	locks the Model while the models are training Used in cancel method More...

std::atomic< bool >	is_joining_
	specifies if a thread for joining the training process has been launched. More...

unsigned int	models_still_training_
	Number of models that are still training. More...

std::mutex	event_mutex_
	Mutex that prevents concurrent calls to onEvent() More...

Detailed Description

Hierarchical Hidden Markov Model for Continuous Recognition and Regression (Multi-class)

Constructor & Destructor Documentation

xmm::HierarchicalHMM::HierarchicalHMM ( bool bimodal = false )

Constructor.

Parameters

bimodal specifies if the model should be used for regression

xmm::HierarchicalHMM::HierarchicalHMM ( HierarchicalHMM const & src )

Copy Constructor.

Parameters

src	Source Model

xmm::HierarchicalHMM::HierarchicalHMM ( Json::Value const & root )

explicit

Constructor from Json Structure.

Parameters

root	Json Value

Member Function Documentation

std::string xmm::Writable::__str__ ( ) const

inlineinherited

"print" method for python => returns the results of write method

Warning: only defined if SWIGPYTHON is defined

void xmm::HierarchicalHMM::addExitPoint	(	int	state,
		float	proba
	)

void xmm::HierarchicalHMM::addModelForClass ( std::string const & label )

protectedvirtual

Update training set for a specific label.

Parameters

label label of the sub-training set to update

Exceptions

out_of_range if the label does not exist

Reimplemented from xmm::Model< SingleClassHMM, HMM >.

void xmm::Model< SingleClassHMM , HMM >::cancelTraining ( )

inlineinherited

Cancels the training of all models.

void xmm::Model< SingleClassHMM , HMM >::cancelTraining ( std::string const & label )

inlineinherited

Cancels the training of a given class.

Parameters

label label of the class to cancel

void xmm::Model< SingleClassHMM , HMM >::checkConfigurationChanges ( ) const

inlineprotectedinherited

Look for configuration changes and throws an exception if the model is not up to date.

void xmm::Model< SingleClassHMM , HMM >::checkTraining ( ) const

inlineprotectedinherited

Checks if the Model is still training.

Exceptions

runtime_error if the Model is training.

void xmm::HierarchicalHMM::clear ( )

virtual

Remove All models.

Reimplemented from xmm::Model< SingleClassHMM, HMM >.

static const double xmm::HierarchicalHMM::DEFAULT_EXITTRANSITION ( )

inlinestatic

Default exit transition for the highest level.

void xmm::HierarchicalHMM::filter ( std::vector< float > const & observation )

virtual

filters a incoming observation (performs recognition or regression)

the results of the inference process are stored in the results attribute

Parameters

observation observation vector

Reimplemented from xmm::Model< SingleClassHMM, HMM >.

void xmm::HierarchicalHMM::forward_init ( std::vector< float > const & observation )

protected

Initialization of the Forward Algorithm for the hierarchical HMM. see: Jules Francoise. Realtime Segmentation and Recognition of Gestures using Hierarchical Markov Models. Master’s Thesis, Université Pierre et Marie Curie, Ircam, 2011. [http://articles.ircam.fr/textes/Francoise11a/index.pdf].

Parameters

observation observation vector. If the model is bimodal, this should be allocated for both modalities, and should contain the observation on the input modality. The predicted output will be appended to the input modality observation

void xmm::HierarchicalHMM::forward_update ( std::vector< float > const & observation )

protected

Update of the Forward Algorithm for the hierarchical HMM. see: Jules Francoise. Realtime Segmentation and Recognition of Gestures using Hierarchical Markov Models. Master’s Thesis, Université Pierre et Marie Curie, Ircam, 2011. [http://articles.ircam.fr/textes/Francoise11a/index.pdf].

Parameters

observation observation vector. If the model is bimodal, this should be allocated for both modalities, and should contain the observation on the input modality. The predicted output will be appended to the input modality observation

void xmm::HierarchicalHMM::fromJson ( Json::Value const & root )

virtual

Read the object from a JSON Structure.

Parameters

root	JSON value containing the object's information

Exceptions

JsonException if the JSON value has a wrong format

Implements xmm::Writable.

int xmm::Model< SingleClassHMM , HMM >::getIndex ( std::string const & label ) const

inlineinherited

Checks if a class exists.

Parameters

label class label

Returns: true if the class labeled 'label' exists

bool xmm::Model< SingleClassHMM , HMM >::hasClass ( std::string const & label ) const

inlineinherited

Checks if a class exists.

Parameters

label class label

Returns: true if the class labeled 'label' exists

void xmm::HierarchicalHMM::joinTraining ( )

protectedvirtual

Finishes the background training process by joining threads and deleting the models which training failed.

: updates transition parameters after joining

Reimplemented from xmm::Model< SingleClassHMM, HMM >.

void xmm::HierarchicalHMM::likelihoodAlpha	(	int	exitNum,
		std::vector< double > &	likelihoodVector
	)		const

protected

get instantaneous likelihood

get instantaneous likelihood on alpha variable for exit state exitNum.

Parameters

exitNum	number of exit state (0=continue, 1=transition, 2=back to root). if -1, get likelihood over all exit states
likelihoodVector	likelihood vector (size nbPrimitives)

void xmm::HierarchicalHMM::normalizeTransitions ( )

protected

Normalize segment level prior and transition matrices.

void xmm::Model< SingleClassHMM , HMM >::onTrainingEvent ( TrainingEvent const & e )

inlineprotectedinherited

Monitors the training of each Model of the group.

xmm::HierarchicalHMM & xmm::HierarchicalHMM::operator= ( HierarchicalHMM const & src )

Assignment.

Parameters

src	Source Model

void xmm::Writable::readFile ( char * fileName )

inlineinherited

read method for python wrapping ('read' keyword forbidden, name has to be different)

Warning: only defined if SWIGPYTHON is defined

void xmm::HierarchicalHMM::removeClass ( std::string const & label )

virtual

Remove a specific class by label.

Parameters

label label of the class to remove

Reimplemented from xmm::Model< SingleClassHMM, HMM >.

void xmm::HierarchicalHMM::reset ( )

virtual

Resets the fitering process (recognition or regression)

Reimplemented from xmm::Model< SingleClassHMM, HMM >.

unsigned int xmm::Model< SingleClassHMM , HMM >::size ( ) const

inlineinherited

Get the number of classes in the model.

Returns: the number of classes in the model

Json::Value xmm::HierarchicalHMM::toJson ( ) const

virtual

Write the object to a JSON Structure.

Returns: Json value containing the object's information

Reimplemented from xmm::Model< SingleClassHMM, HMM >.

virtual void xmm::Model< SingleClassHMM , HMM >::train ( TrainingSet * trainingSet )

inlinevirtualinherited

Train all classes from the training set passed in argument.

Parameters

trainingSet Training Set

virtual void xmm::Model< SingleClassHMM , HMM >::train	(	TrainingSet *	trainingSet,
		std::string const &	label
	)

inlinevirtualinherited

Train a specific class from the training set passed in argument.

Parameters

trainingSet	Training Set
label	label of the class to train

Exceptions

out_of_range	if the label does not exist
runtime_error	if the dimensions of the training set don't match the dimensions of the current model

bool xmm::Model< SingleClassHMM , HMM >::trained ( ) const

inlineinherited

Checks if the model is trained (training finished and not empty)

bool xmm::Model< SingleClassHMM , HMM >::training ( ) const

inlineinherited

Checks if the model is still training.

void xmm::HierarchicalHMM::updateExitProbabilities ( )

protected

Update exit probabilities of each sub-model.

void xmm::HierarchicalHMM::updatePrior ( )

protected

ergodic learning update high-level prior probabilities -> equal prior probs

void xmm::HierarchicalHMM::updateResults ( )

protected

Update the results (Likelihoods)

void xmm::HierarchicalHMM::updateTransition ( )

protected

ergodic learning: update high-level transition matrix

equal transition probabilities between primitive gestures

void xmm::HierarchicalHMM::updateTransitionParameters ( )

protected

update high-level parameters when a new primitive is learned

updated parameters: prior probabilities + transition matrix

void xmm::Writable::writeFile ( char * fileName ) const

inlineinherited

write method for python wrapping ('write' keyword forbidden, name has to be different)

Warning: only defined if SWIGPYTHON is defined

Member Data Documentation

std::atomic<bool> xmm::Model< SingleClassHMM , HMM >::cancel_required_

protectedinherited

locks the Model while the models are training Used in cancel method

Configuration<HMM > xmm::Model< SingleClassHMM , HMM >::configuration

inherited

Configuration (default and class-specific parameters)

std::mutex xmm::Model< SingleClassHMM , HMM >::event_mutex_

protectedinherited

Mutex that prevents concurrent calls to onEvent()

std::vector<double> xmm::HierarchicalHMM::exit_transition

exit probabilities of the model (probability to finish and go back to the root)

bool xmm::HierarchicalHMM::forward_initialized_

protected

Defines if the forward algorithm has been initialized.

std::vector<double> xmm::HierarchicalHMM::frontier_v1_

protected

intermediate Forward variable (used in Frontier algorithm)

std::vector<double> xmm::HierarchicalHMM::frontier_v2_

protected

intermediate Forward variable (used in Frontier algorithm)

std::atomic<bool> xmm::Model< SingleClassHMM , HMM >::is_joining_

protectedinherited

specifies if a thread for joining the training process has been launched.

std::atomic<bool> xmm::Model< SingleClassHMM , HMM >::is_training_

protectedinherited

locks the Model while the models are training Used in cancel method

std::map<std::string, SingleClassHMM > xmm::Model< SingleClassHMM , HMM >::models

inherited

models stored in a map. Each Model is associated with a label

unsigned int xmm::Model< SingleClassHMM , HMM >::models_still_training_

protectedinherited

Number of models that are still training.

std::vector<double> xmm::HierarchicalHMM::prior

Prior probabilities of the models.

Results<HMM> xmm::HierarchicalHMM::results

Convert to bimodal HierarchicalHMM in place.

Parameters

dimension_input dimension of the input modality

Exceptions

runtime_error	if the model is already bimodal
out_of_range	if the requested input dimension is too large Convert to unimodal HierarchicalHMM in place
runtime_error	if the model is already unimodal extract a submodel with the given columns

Parameters

columns columns indices in the target order

Exceptions

runtime_error	if the model is training
out_of_range	if the number or indices of the requested columns exceeds the current dimension

Returns: a HierarchicalHMM from the current model considering only the target columns extract the submodel of the input modality

Exceptions

runtime_error if the model is training or if it is not bimodal

Returns: a unimodal GMM of the input modality from the current bimodal model extract the submodel of the output modality

Exceptions

runtime_error if the model is training or if it is not bimodal

Returns: a unimodal HierarchicalHMM of the output modality from the current bimodal model extract the model with reversed input and output modalities

Exceptions

runtime_error if the model is training or if it is not bimodal

Returns: a bimodal HierarchicalHMM that swaps the input and output modalities Results of the Filtering Process (Recognition + Regression)

std::shared_ptr<SharedParameters> xmm::Model< SingleClassHMM , HMM >::shared_parameters

inherited

Set of Parameters shared among classes.

EventGenerator<TrainingEvent> xmm::Model< SingleClassHMM , HMM >::training_events

inherited

Generator for training process events.

std::map<std::string, std::thread> xmm::Model< SingleClassHMM , HMM >::training_threads_

protectedinherited

Training Threads.

std::vector<std::vector<double> > xmm::HierarchicalHMM::transition

Transition probabilities between models.

The documentation for this class was generated from the following files:

XMM - Probabilistic Models for Motion Recognition and Mapping

Public Member Functions

Static Public Member Functions

Public Attributes

Protected Member Functions

Protected Attributes

Detailed Description

Constructor & Destructor Documentation

Member Function Documentation

Member Data Documentation