[試題] 98上 李琳山 語音處理概論 期末考
課程名稱︰語音處理概論
課程性質︰選修
課程教師︰李琳山 教授
開課學院：電資學院
開課系所︰電機系
考試日期（年月日）︰2010/01/15
考試時限（分鐘）：120
是否需發放獎勵金：
（如未明確表示，則不予發放）
試題 :
1.In feature based approach of robust speech recognition, we mentioned Cepstral
Mean Subtraction(CMS), Cepstral Mean and Variance Normalization()CMVN), and
Histogram Equalization(HEQ). For each of them, please explain.
(a) what they are, and
(b) why they work
Explain why HEQ can outperform CMS and CMVN
2.What is Parallel Model Combination(PMC) approach for model-based robust
speech recognition? Explain how it works.
3.In a spoken document retrieval system, many queries are out of vocabulary
(OOV) words. Explain why OOV words are problems and how this problem is
handled?
4.In eigenvoice approach, explain how the eigenvoice space is constructed, what
that means, and why rapid speaker adaptation can be achieved with very
limited quantity of data?
5.In Latent Semantic Analysis(LSA), the element w_ij of the word-document
matrix W is
w_i=(1-ε_ij)*c_ij/n_j
where c_ij 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.
6.What is the goal of Principal Component analysis? Write down the procedure
for PCA.
--
※ 發信站: 批踢踢實業坊(ptt.cc)
◆ From: 140.112.242.115
推
07/28 14:51, , 1F
07/28 14:51, 1F