D i g i t a l S i g n a l P ro c e s s i n g 英 文 试 卷 样 卷
该试卷从吴镇杨老师编写《数字信号处理》参考
书后翻译，不代表真实考试难度，仅做题型参考
课程名称 数字信号处理 考试学期 Sample 得分
适用专业 信息工程 考试形式 半开卷 考试时间长度 120 分钟
开卷、半开卷可带的资料：
（4 页单面或 2 页双面 A4 纸）
Section A (10%): True or False （1 mark per question，10 marks in total)
1. For FIR filters designed with Kaiser window, the transition band gets narrower
with reducing β. ( )
自 2. The poles of all-pass system must be within the unit circle, and its zeros are
觉 conjugate reciprocals of the poles. ( )
遵 3. If we have 𝐻1 (𝑧) = 𝐻2 (𝑧) , we must have ℎ1 [𝑛] = ℎ2 [𝑛] , where ℎ1 [𝑛] and
守
ℎ2 [𝑛] are the inverse z-transform of 𝐻1 (𝑧) and 𝐻2 (𝑧), respectively. ( )
考
4. IIR System must have recursive structure, i.e., feedback from the output. ( )
场
5. Direct form structure of digital IIR filters is less sensitive to quantization effect
纪
of coefficients, as compared with the cascade and parallel structures. ( )
线
律
6. The sampling points of Chirp-z transform do not have to locate on the unit
circle. ( )
如 7. With sampling frequency 𝑓𝑠 = 5000 Hz, 2000-point DFT samples the
姓名
考 spectrum with frequency interval 2.5 Hz. ( )
试
8. A N -point sequence can be uniquely determined by N samples from its
作
封
frequency spectrum. ( )
弊
9. Equiripple approximation (Parks-McClellan algorithm) is an optimal FIR filter
此 design strategy that minimize the maximum error. ( )
答 10. The overflow oscillation is caused by round off noise. ( )
卷
密
无 Section B (10%): Multiple Choice（1 mark per question，10 marks in total)
效 1. To evaluate the linear convolution of two N-point sequences using DFT algorithms,
what is the minimum length for the DFT computations?
学号
a. N b. 2N-1 c. 2N+1
共 4 页 第 1 页
2. Which of the following IIR filter has ripple for both passband and stopband?
a. Elliptic b. Chebyshev c. Butterworth
3. Which of the following settings on ℎ[𝑛], 0 ≤ 𝑛 ≤ 𝑁 − 1 , can be used for
designing a linear phase lowpass FIR filter?
a. h[n] is even symmetric, N is even b. h[n] is odd symmetric, N is odd
c. h[n] is odd symmetric, N is even
4. Consider a sequence x[n] with sampling frequency 𝑓𝑠 and its DFT represented
by X[k], 0 ≤ k ≤ N − 1. The value X[2] represents the frequency sample at ( ).
a. 𝑓𝑠 /𝑁 b. 2𝑓𝑠 /𝑁 c. 𝑓𝑠 /2𝑁
5. How is sequence x[2n] related with sequence x[n]?
a. x[2n] is obtained from x[n] by dividing its amplitude by 2
b. x[2n] is obtained from x[n] by multiplying its amplitude with 2
c. x[2n] is obtained from x[n] by extracting the sample points with even indices.
6. A system whose output depends only on the past and current inputs, but not the
future inputs, is
a. stable system b. causal system c. time-invariant system
(𝑛−𝑝)2
7. Taking 16 point FFT of the Gaussian sequence 𝑥[𝑛] = exp[− ]. Which of
𝑞
the following parameters will lead to the most significant low frequency component?
a. 𝑝 = 8, 𝑞 = 2 b. 𝑝 = 8, 𝑞 = 8 c. 𝑝 = 14, 𝑞 = 8
8. Consider a digital filter, taking the DTFT of its impulse response gives:
a. its power spectrum b. its frequency response c. its amplitude spectrum.
9 . The quantization noise of the Nth order FIR filter is____________.
a. proportional to N， b. proportional to N2， c. proportional to N3.
10. In general, the ________ of the decimation-in-time FFT algorithm is in bit-reversal
order.
a. input b. output c. input and output
共 4 页 第 2 页
Section C (32%): Calculations（8 marks per question，32 marks in total)
1. Calculate the output of the following MATLAB codes:
2. If 𝐻(𝑧) is a low pass filter, which type of filter does 𝐻(−𝑧) belong to? Let 𝜔𝑝
and 𝜔𝑠 denote the passband and stopband edge frequency, respectively. Find the
passband and stopband edge frequency of 𝐻(−𝑧) and the relationship between
the impulse response of these two filters.
3. Consider the analog signal
xa(t)=3cos(2000πt)+5sin(6000πt)+10cos(12000πt)
(a) What is the Nyquist rate for this signal？
(b) If we sample this signal using a sampling rate fs=5000Hz. What is the discrete-time
signal obtained after sampling？
4. Design a 3rd order digital Butterworth filter using bilinear transformation method,
with sampling frequency 9𝑘𝐻𝑧 and passband cutoff frequency 3𝑘𝐻𝑧.
Note: the transfer function of the 3rd order digital Butterworth filter prototype is
1
Ha s
1 2 s / c 2 s / c s / c
2 3
Section D (48%): Problems（12 marks per question，48 marks in total)
1. Consider a finite-length complex sequence 𝑓[𝑛] of length 𝑁 , which is
constructed by a real sequence 𝑥[𝑛] of length 2𝑁, i.e., 𝑓[𝑛] = 𝑥[2𝑛] + 𝑗𝑥[2𝑛 +
1]. Denote the DFT of 𝑓[𝑛] and 𝑥[𝑛] by 𝐹[𝑘] and 𝑋[𝑘], respectively. Find the
expression of 𝑋[𝑘] in terms of 𝐹[𝑘].
共 4 页 第 3 页
2. Find the best cascade structure of the filter with system function 𝐻(𝑧) =
1
， such that the round-off noise at the output is minimized, and
(1−0.6𝑧 −1 )(1−0.9𝑧 −1 )
calculate the round-off noise variance at the output.
3. Use frequency sampling method to design a linear-phase highpass FIR digital filter,
with ideal frequency response
0, 0 0.8
H d e j
1, 0.8
Take N=16 sample points and give the sample values 𝐻[𝑘] and the impulse response
ℎ[𝑛].
4. x[n] is a 8-point real sequence, X k X k e j X k
DFT x n and the first 5
points in X[k] are 4, 2 3 j, 1 2 j, 4 j, 2 ；y[n] is obtained from x[n] by
circular shift of 4 positions; Y k Y k e DFT y n .
jY k
7 7
x[n] , x[n]
2
a). Find x[0], 。
n0 n0
(2) Is X [k ] Y [k ] true? How is x k related with y k ?
共 4 页 第 4 页
