Reconocimiento del habla

Temas a tratar en el curso:
Los temas a tratar en el curso (utilizando material proveniente del curso de reconocimiento de voz desarrollado por Microsoft son:
- Conocimientos previos y bases matemáticas
- Procesamiento de señales de voz
- Modelado acústico
- Modelado de lenguajes
- Decodificación del habla
- Modelado acústico avanzado
Conocimientos previos
- Fonética (definiciones básicas)
- Medidas de desempeño
- Ecuación fudamental del reconocimiento del habla
Pequeño problemario
- Las conosonates se caracterizan por:
a) the vibration of the vocal chords at a particular frequency
b) significant constriction of air flow in the airway or mouth
c) Loud volume compared to vowels
d) Number of syllables
- La pronunciación fonética n ey sh ah n corresponde a la palabra (en inglés):
a) Nasal
b) Nathan
c) Nation
d) Nissan
-
Assume that 1% of the population are musicians and that 10% of the total population is left-handed. A recent survey of musicians reveals that 60% of them are left-handed. What is the probability that a left-handed toddler will be a musician?
-
What would be the new probability if 50% of the population was left-handed?
-
What is the name for the atomic unit of speech sound?
-
Which of these is a common measure of statistical significance for speech recognition?
a) Student-t Test
b) Chi-Squared Test
c) Matched Pairs Test
d) Binomial Test
- The fundamental equation of speech recognition uses which mathematical rule?
a) L’Hopital’s Rule
b) Chi-Squared Test
c) Bayes’ Rule
d) Binomial Test
- Which of these can be used to measure the performance of a Speech Recognition System?
a) Word Error Rate
b) Sentence Error Rate
c) Latency
d) All of the Above
- In the fundamental equation of Speech Recognition, P(W) represents the…
a) Acoustic Model
b) Language Model
c) Pronunciation Model
d) None of the Above
-
In the fundamental equation of Speech Recognition, P(O W) represents the… (Choose One)
a) Acoustic Model
b) Language Model
c) Pronunciation Model
d) None of the Above