课程名称︰数位语音处理概论
课程性质︰选修
课程教师︰李琳山
开课学院：电资学院
开课系所︰电机、资工系
考试日期（年月日）︰2008.06.17
考试时限（分钟）：120
是否需发放奖励金：是
（如未明确表示，则不予发放）
试题 :
Digital Speech Processing, Final Exam
June. 17, 2008, 10:10-12:10
● OPEN Lecture Notes, Power Point(Printed Version) and Personal Notes.
● You have to use CHINESE sentences to answer all of the problems, but you can
use English terminology.
● Total points: 120
● Note that you don't need to be able to answer all the questions.
───────────────────────────────────────
1. (20) In Maximum Likelihood Linear Regression (MLLR) approach, explain how
the mean vectors of Gaussian mixtures in acoustic models are adjusted, why
speaker adaptation can be achieved with relatively smaller quantity of data,
and what are the limitations?
2. (20) In Latent Semantic Analysis the elements w_ij of the word-document
_
matrix W is
c_ij
w_ij = (1 - ε_i) * ─── ,
n_j
Where c_ij is the number of times the word w_i occurs in the document d_j,
n_j is the total number of words in d_j, and
1 N ┌ c_ij ┐ ┌ c_ij ┐ N
ε_i = - ─── * Σ│───│log│───│ , t_i = Σ c_ij ,
log(N) j=1└ t_i ┘ └ t_i ┘ j=1
where N is the total number of documents. Explain the meaning of the
parameters w_ij, and the meaning of each row and column of this matrix.
3. (20) Explain why subword units are useful in retrieving speech information
using speech queries, and if they introduce extra problems?
4. (20) For Linear Discriminative Analysis (LDA), explain the meaning of
within/between-class scatter matrices S_W/S_B, and the meaning of the
optimization criterion.
5. (20) Write down the two steps in each iteration of EM algorithm and explain
how they operate and what they mean.
6. (20) Explain the three key elements in a spoken dialogue system: what they
are, how they operate, and how they are linked together.