tratamiento digital de voz prof. luis a. hernández gómez ftp.gaps.ssr.upm.es/pub/tdv/doc/...

11
Tratamiento Digital de Voz Prof. Luis A. Hernández Gómez ftp.gaps.ssr.upm.es/pub/TDV/D OC/ Tema2c.ppt Dpto. Señales, Sistemas Dpto. Señales, Sistemas y Radiocomunicaciones y Radiocomunicaciones

Upload: damian-ford

Post on 30-Dec-2015

218 views

Category:

Documents


1 download

TRANSCRIPT

Page 1: Tratamiento Digital de Voz Prof. Luis A. Hernández Gómez ftp.gaps.ssr.upm.es/pub/TDV/DOC/ Tema2c.ppt Dpto. Señales, Sistemas y Radiocomunicaciones

Tratamiento Digital de Voz

Prof. Luis A. Hernández Gómez

ftp.gaps.ssr.upm.es/pub/TDV/DOC/Tema2c.ppt

Dpto. Señales, Sistemas y Dpto. Señales, Sistemas y RadiocomunicacionesRadiocomunicaciones

Page 2: Tratamiento Digital de Voz Prof. Luis A. Hernández Gómez ftp.gaps.ssr.upm.es/pub/TDV/DOC/ Tema2c.ppt Dpto. Señales, Sistemas y Radiocomunicaciones

Agenda• Perceptual evaluation of Speech Quality• Traditional evaluation of Speech Quality• A new approach to evaluation of Speech Quality• Perceptual Evaluation of Speech Quality within

ITU-T

Page 3: Tratamiento Digital de Voz Prof. Luis A. Hernández Gómez ftp.gaps.ssr.upm.es/pub/TDV/DOC/ Tema2c.ppt Dpto. Señales, Sistemas y Radiocomunicaciones

The idea

• To simulate the sound perception of subjects in real-life situations

• To have a objective technique based on a perceptual model that will reveal the same MOS score as that of Listening Test

What is MOS?

• Methods and procedures for conducting subjective evaluation of transmission quality (ITU -T Rec. 800)

• Quality of speech based on Listening Tests

A new approach to evaluation of Speech Quality

Page 4: Tratamiento Digital de Voz Prof. Luis A. Hernández Gómez ftp.gaps.ssr.upm.es/pub/TDV/DOC/ Tema2c.ppt Dpto. Señales, Sistemas y Radiocomunicaciones

Objective techniques used for predicting subjective test scores

• PSQM (Perceptual Speech Quality Measure)

• PSQM+ (ITU -T Rec. 861)

• PESQ (Perceptual Evaluation of Speech Quality) (ITU -T Rec. 862)

• and others like MNB, PAMS, TOSQA, PACE, VQI and PESQM.

About PSQM, PSQM+, PESQ

• For jugging the listening and talking quality of Telephone band speech signals (300-3400Hz)

• Signals of the input and output of the device under test are mapped onto a psychophysical representation that match as close as possible the internal representation inside our head

A new approach to evaluation of Speech Quality

Page 5: Tratamiento Digital de Voz Prof. Luis A. Hernández Gómez ftp.gaps.ssr.upm.es/pub/TDV/DOC/ Tema2c.ppt Dpto. Señales, Sistemas y Radiocomunicaciones

Agenda• Perceptual evaluation of Speech Quality• Traditional evaluation of Speech Quality• A new approach to evaluation of Speech Quality• Perceptual Evaluation of Speech Quality within

ITU-T

Page 6: Tratamiento Digital de Voz Prof. Luis A. Hernández Gómez ftp.gaps.ssr.upm.es/pub/TDV/DOC/ Tema2c.ppt Dpto. Señales, Sistemas y Radiocomunicaciones

Generic perceptual measurement algorithm

Perceptual Model

Feature-Extractor

Test

Perceptual Model

Reference

Cognitive Model

Quality measure

Impairment Grade

Excellent 5

Good 4

Fair 3

Poor 2

Bad 1

Perceptual Model

Is a model of the Human Ear

Cognitive Model

Is a model of the judgement behaviour

of the test subject

Page 7: Tratamiento Digital de Voz Prof. Luis A. Hernández Gómez ftp.gaps.ssr.upm.es/pub/TDV/DOC/ Tema2c.ppt Dpto. Señales, Sistemas y Radiocomunicaciones

Basic model of PSQM & PSQM+ algorithm

Perceptual Model

Internal representation of the reference signal

Test

Perceptual Model

Reference

Cognitive Model

Quality measureDifference in internal

representation

Internal representation of the test signal

Improvements in PSQM+• Time alignment: variable delay, frame repeats• Weight of distortion: time clipping, time frequency distortion

Improvements in PSQM+

• Time alignment: variable delay, frame repeats• Weight of distortion: time clipping, time frequency distortion

Page 8: Tratamiento Digital de Voz Prof. Luis A. Hernández Gómez ftp.gaps.ssr.upm.es/pub/TDV/DOC/ Tema2c.ppt Dpto. Señales, Sistemas y Radiocomunicaciones

Basic model of PESQ algorithm

Auditory transform

Degraded

Level align

Reference

Cognitive Modelling

Quality measure

Input filter

Auditory transform

Disturbance processing

Identify bad intervals

Time align and

equalise

Level align

Input filter

Next step is PESQM

• Assessment of Handset on a perceptual basis using HATS

• Covering echo and Sidetone

Next step is PESQM

• Assessment of Handset on a perceptual basis using HATS

• Covering echo and Sidetone

Page 9: Tratamiento Digital de Voz Prof. Luis A. Hernández Gómez ftp.gaps.ssr.upm.es/pub/TDV/DOC/ Tema2c.ppt Dpto. Señales, Sistemas y Radiocomunicaciones

Perceptual measurement algorithms - roadmap

PSQM P.861, 1996

Intrusive Narrowband

Speech Quality

PSQM+ 1996

PESQ P.862, 1996

PESQM 2001,2002

Conversational Quality

P3SQM 2001,2002

Echo and Acoustical

measurement

Non-Intrusive Narrowband

Speech Quality

Wideband Audio

Video measurement

PEAQ BS1387, 1996

PEAQ+ 2002

Acoustical PESQ

extension

Single-ended Voice Quality

PEVQ

Wideband Voice Quality

Audiovisual Quality

Page 10: Tratamiento Digital de Voz Prof. Luis A. Hernández Gómez ftp.gaps.ssr.upm.es/pub/TDV/DOC/ Tema2c.ppt Dpto. Señales, Sistemas y Radiocomunicaciones
Page 11: Tratamiento Digital de Voz Prof. Luis A. Hernández Gómez ftp.gaps.ssr.upm.es/pub/TDV/DOC/ Tema2c.ppt Dpto. Señales, Sistemas y Radiocomunicaciones