Perceptron Class Reference

The perceptron. More...

#include <Perc.h>

List of all members.

Public Member Functions
	Perceptron ()
	Creates a new perceptron with default time specifics, an empty frequency mask, without hysteresis.
	~Perceptron ()
	Destroys the perceptron.
void	RandomVector ()
	Randomly extracts a radius and generates the separation surface of the perceptron as an hypersphere.
int	Load (const char *filename)
	Loads the perceptron from a file.
int	Save (const char *filename)
	Saves the perceptron.
int	Learn (ExampleType *Example, int numExamples, float learningRate)
	Executes a learning cycle on the perceptron.
int	Exec (float *Input)
	Calculates the output of the perceptron.
void	SetFrequencyMask (FrequencyMaskType fMask, bool useDoppler)
	Sets the percpetron's frequency mask.
void	SetTurnOnThreshold (float s)
	Sets the Noise->Whistle (turn-on) threshold.
void	SetTurnOffThreshold (float s)
	Sets the Whistle->Noise (turn-off) threshold.
void	SetTimes (TimeSpecsType)
	Sets perceptron's time specifics.
TimeSpecsType	GetTimes ()
	Gets the perceptron's time specifics.
FrequencyMaskType	GetFrequencyMask ()
	Returns the perceptron's frequency mask.
float	GetTurnOnThreshold ()
	Returns the perceptron's Noise->Whistle (turn-on) threshold, in dB.
float	GetTurnOffThreshold ()
	Returns the perceptron's Whistle->Noise (turn-off) threshold, in dB.
bool	UseDoppler ()
	Tells if the perceptron is using doppler compensation.
Public Attributes
int	mMicLevel
	IGAIN value used during recognition.
Private Attributes
int	mState
	status of the perceptron : +1=Whistle, -1=Noise
float	mTurnOnThreshold
	Noise->Whistle (turn-on) threshold.
float	mTurnOffThreshold
	Whistle->Noise (turn-off) threshold.
FrequencyMaskType	mFmask
	standard frequency mask
FrequencyMaskType	mExtFmask
	extended frequency mask (including doppler compensation)
float	mW [32]
	Weigth vector.
bool	mUseDoppler
	Doppler compensation.
TimeSpecsType	mTimeSpecs
	Time specifics for the whistle counter.

---

Detailed Description

The perceptron.

---

Constructor & Destructor Documentation

Perceptron::Perceptron (   )

Creates a new perceptron with default time specifics, an empty frequency mask, without hysteresis.

Perceptron::~Perceptron (   )

Destroys the perceptron.

---

Member Function Documentation

int Perceptron::Exec (  float *  Input  )

Calculates the output of the perceptron. Parameters: [in]  Input  The input sequence, obtained as the FrequencyMask-ed periodogram of the input audio sequence Returns: +1:Whistle -1:Noise

FrequencyMaskType Perceptron::GetFrequencyMask (   )

Returns the perceptron's frequency mask. Returns: The frequency mask

TimeSpecsType Perceptron::GetTimes (   )

Gets the perceptron's time specifics. Returns: Time specifics the perceptron is using

float Perceptron::GetTurnOffThreshold (   )

Returns the perceptron's Whistle->Noise (turn-off) threshold, in dB. Returns: The turn-off threshold

float Perceptron::GetTurnOnThreshold (   )

Returns the perceptron's Noise->Whistle (turn-on) threshold, in dB. Returns: The turn-on threshold

int Perceptron::Learn (  ExampleType *  Example, int  numExamples, float  learningRate )

Executes a learning cycle on the perceptron. Parameters: [in]  Example  Pointer to the first example numExamples  Numeber of examples learningRate  Learning rate to be used in this learning cycle Returns: This function returns the number of mispredictions during this cycle

int Perceptron::Load (  const char *  filename  )

Loads the perceptron from a file. Parameters: filename  The name of the perceptron file Returns: 1:succes -1:error - use Perc_GetLastError() for details A fac-simile of a perceptron file follows: time_step= 20 short_steps= 25 long_steps= 50 interval_steps= 25 min_whistle_steps= 6 min_noise_steps= 6 W 0= 352.229 W 1= 352.229 W 2= 352.229 W 3= 352.229 W 4= 352.229 W 5= 352.229 W 6= 352.229 W 7= 352.229 W 8= 352.229 W 9= 352.229 W10= 352.229 W11= 352.229 W12= 558.618 W13= 48.161 W14= 1310.996 W15= 352.229 W16= 352.229 W17= 352.229 W18= 352.229 W19= 352.229 W20= 352.229 W21= 352.229 W22= 352.229 W23= 352.229 W24= 352.229 W25= 352.229 W26= 352.229 W27= 352.229 W28= 352.229 W29= 352.229 W30= 352.229 W31= 352.229 turn_on_threshold= 0.100 turn_off_threshold= -1.500 frequency_mask= 00000000000011100000000000000000 doppler_compensation= 1 mic_level= 100

void Perceptron::RandomVector (   )

Randomly extracts a radius and generates the separation surface of the perceptron as an hypersphere.

int Perceptron::Save (  const char *  filename  )

Saves the perceptron. Parameters: filename  The file which will contain the perceptron Returns: 1:success -1:error - use Perc_GetLastError() for details

void Perceptron::SetFrequencyMask (  FrequencyMaskType  fMask, bool  useDoppler )

Sets the percpetron's frequency mask. Parameters: [in]  fMask:  the frequency mask useDoppler:  If true, perceptron's frequency mask is extended as described in the documentation (Chapter 1)

void Perceptron::SetTimes (  TimeSpecsType   )

Sets perceptron's time specifics. Parameters: [in]  tspecs  Time specifics to be used.

void Perceptron::SetTurnOffThreshold (  float  s  )

Sets the Whistle->Noise (turn-off) threshold. Parameters: s  The threshold referred to the separation surface, expressed in dB

void Perceptron::SetTurnOnThreshold (  float  s  )

Sets the Noise->Whistle (turn-on) threshold. Parameters: s  The threshold referred to the separation surface, expressed in dB

bool Perceptron::UseDoppler (   )

Tells if the perceptron is using doppler compensation. Returns: true if the perceptron is using doppler compensation false otherwise

---

Member Data Documentation

FrequencyMaskType Perceptron::mExtFmask [private]

extended frequency mask (including doppler compensation)

FrequencyMaskType Perceptron::mFmask [private]

standard frequency mask

int Perceptron::mMicLevel

IGAIN value used during recognition.

int Perceptron::mState [private]

status of the perceptron : +1=Whistle, -1=Noise

TimeSpecsType Perceptron::mTimeSpecs [private]

Time specifics for the whistle counter.

float Perceptron::mTurnOffThreshold [private]

Whistle->Noise (turn-off) threshold.

float Perceptron::mTurnOnThreshold [private]

Noise->Whistle (turn-on) threshold.

bool Perceptron::mUseDoppler [private]

Doppler compensation.

float Perceptron::mW[32] [private]

Weigth vector.

---

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

• Perc.h

---