class SpeechHMM

This class implements a special case of Hidden Markov Models that can be used to do connected word speech recognition for small vocabulary, using embedded training.

Inheritance:

Public Fields

int n_models

the number of basic phoneme models

HMM** models

the basic phoneme models

DataSet* data

a dataset for initialization

EMTrainer** model_trainer

if an initial alignment is given and an emtrainer for each model then it is used to train the models after kmeans during reset

MeasurerList* initial_models_trainer_measurers

as well as a measurer of this trainer

LexiconInfo* lexicon

the acceptable lexicon

Sequence* targets

the current target sequence, with start and end words/phonemes

int add_to_targets

number of words to add

bool** word_transitions

true if the given transition is a transition between words

int max_n_states

the maximum number of states in the graph (used for allocation)

int* states_to_model_states

the relation between model states and SpeechHMM states

int* states_to_model

the relation between models and SpeechHMM states

int* states_to_word

the relation between words and SpeechHMM states

bool phoneme_targets

are targets expressed in words or phonemes?

Public Methods

SpeechHMM(int n_models_, HMM** models_, LexiconInfo* lex_, EMTrainer** model_trainer_ = NULL)

In order to create a SpeechHMM, we need to give a vector of n_models_ HMMs as well as their corresponding name, a lexicon, an optional log_word_entrance_penalty and an optional trainer that can be used to initialize each model independently

virtual void prepareTrainModel(Sequence* input)

this method prepare the transition graph associated with a given training sentence

virtual int addWordToModel(int word, int current_state)

this method is used by prepareTrainModel to prepare the model.

virtual void addConnectionsBetweenWordsToModel(int word, int next_word, int current_state, int next_current_state, real log_n_next)

this method is used by prepareTrainModel to prepare the model.

virtual int nStatesInWord(int word)

this methods returns the number of states in a given word

virtual int nStatesInWordPronunciation(int word, int pronun)

this methods returns the number of states in a given word pronunciation

Inherited from HMM:

Public Fields

int n_states

real prior_transitions

Distribution** states

Distribution** shared_states

real** transitions

real** log_transitions

real** dlog_transitions

real** transitions_acc

Sequence* log_alpha

Sequence* log_beta

Sequence* arg_viterbi

int last_arg_viterbi

Sequence* viterbi_sequence

Sequence* log_probabilities_s

bool initialize

Public Methods

virtual void printTransitions(bool real_values=false, bool transitions_only=false)

virtual void logAlpha(Sequence* inputs)

virtual void logBeta(Sequence* inputs)

virtual void logViterbi(Sequence* inputs)

virtual void decode(Sequence* input)

virtual void logProbabilities(Sequence* inputs)

Inherited from Distribution:

Public Fields

real log_probability

Sequence* log_probabilities

Public Methods

virtual real logProbability(Sequence* inputs)

virtual real viterbiLogProbability(Sequence* inputs)

virtual real frameLogProbability(int t, real* f_inputs)

virtual real viterbiFrameLogProbability(int t, real* f_inputs)

virtual void eMIterInitialize()

virtual void iterInitialize()

virtual void eMSequenceInitialize(Sequence* inputs)

virtual void sequenceInitialize(Sequence* inputs)

virtual void eMAccPosteriors(Sequence* inputs, real log_posterior)

virtual void frameEMAccPosteriors(int t, real* f_inputs, real log_posterior)

virtual void viterbiAccPosteriors(Sequence* inputs, real log_posterior)

virtual void frameViterbiAccPosteriors(int t, real* f_inputs, real log_posterior)

virtual void eMUpdate()

virtual void update()

virtual void eMForward(Sequence* inputs)

virtual void viterbiForward(Sequence* inputs)

virtual void frameBackward(int t, real* f_inputs, real* beta_, real* f_outputs, real* alpha_)

virtual void viterbiBackward(Sequence* inputs, Sequence* alpha)

virtual void frameDecision(int t, real* decision)

Public Members

Returns the decision of the distribution

Inherited from GradientMachine:

Public Fields

int n_inputs

int n_outputs

Parameters* params

Parameters* der_params

Sequence* beta

Public Methods

virtual void forward(Sequence* inputs)

virtual void backward(Sequence* inputs, Sequence* alpha)

virtual void setPartialBackprop(bool flag=true)

virtual void frameForward(int t, real* f_inputs, real* f_outputs)

virtual void loadXFile(XFile* file)

virtual void saveXFile(XFile* file)

Inherited from Machine:

Public Fields

Sequence* outputs

Public Methods

virtual void reset()

virtual void setDataSet(DataSet* dataset_)

Inherited from Object:

Public Fields

Allocator* allocator

Public Methods

void addOption(const char* name, int size, void* ptr, const char* help="")

void addIOption(const char* name, int* ptr, int init_value, const char* help="")

void addROption(const char* name, real* ptr, real init_value, const char* help="")

void addBOption(const char* name, bool* ptr, bool init_value, const char* help="")

void addOOption(const char* name, Object** ptr, Object* init_value, const char* help="")

void setOption(const char* name, void* ptr)

void setIOption(const char* name, int option)

void setROption(const char* name, real option)

void setBOption(const char* name, bool option)

void setOOption(const char* name, Object* option)

void load(const char* filename)

void save(const char* filename)

void* operator new(size_t size, Allocator* allocator_=NULL)

void* operator new(size_t size, Allocator* allocator_, void* ptr_)

void operator delete(void* ptr)

Documentation

This class implements a special case of Hidden Markov Models that can be used to do connected word speech recognition for small vocabulary, using embedded training.

It contains a set of phoneme models (represented by HMMs), a lexicon of words (which are sequences of phonemes)

int n_models

the number of basic phoneme models

HMM** models

the basic phoneme models

DataSet* data

a dataset for initialization

EMTrainer** model_trainer

if an initial alignment is given and an emtrainer for each model then it is used to train the models after kmeans during reset

MeasurerList* initial_models_trainer_measurers

as well as a measurer of this trainer

LexiconInfo* lexicon

the acceptable lexicon

Sequence* targets

the current target sequence, with start and end words/phonemes

int add_to_targets

number of words to add

bool** word_transitions

true if the given transition is a transition between words

int max_n_states

the maximum number of states in the graph (used for allocation)

int* states_to_model_states

the relation between model states and SpeechHMM states

int* states_to_model

the relation between models and SpeechHMM states

int* states_to_word

the relation between words and SpeechHMM states

bool phoneme_targets

are targets expressed in words or phonemes?

SpeechHMM(int n_models_, HMM** models_, LexiconInfo* lex_, EMTrainer** model_trainer_ = NULL)

In order to create a SpeechHMM, we need to give a vector of n_models_ HMMs as well as their corresponding name, a lexicon, an optional log_word_entrance_penalty and an optional trainer that can be used to initialize each model independently

virtual void prepareTrainModel(Sequence* input)

this method prepare the transition graph associated with a given training sentence

virtual int addWordToModel(int word, int current_state)

this method is used by prepareTrainModel to prepare the model. It adds a given word to the current graph.

virtual void addConnectionsBetweenWordsToModel(int word, int next_word, int current_state, int next_current_state, real log_n_next)

this method is used by prepareTrainModel to prepare the model. It adds the connections between words.

virtual int nStatesInWord(int word)

this methods returns the number of states in a given word

virtual int nStatesInWordPronunciation(int word, int pronun)

this methods returns the number of states in a given word pronunciation

Direct child classes:

SimpleDecoderSpeechHMM

Author:

Samy Bengio (bengio@idiap.ch)

Alphabetic index HTML hierarchy of classes or Java