SlideShare a Scribd company logo
1 of 19
Learning the MMSE Channel Estimator
EESC 6353 Project
Shamman Noor
Introduction
• System Model
• MMSE Estimation:
- Discrete MMSE estimation
- Structured MMSE estimation
- Circulant structure
- Toeplitz structure
- Fast MMSE estimation
- CNN based fast MMSE estimation
• CNN Model and parameters
Software used:
• MATLAB – 2018a
• Python – 3.7
• TensorFlow – 1.13.1
System Model
…
Base station
(M antennas)
UE
(single antenna)
…
𝑦𝑡 = ℎ 𝑡 + 𝑧𝑡
ℎ1,1
ℎ1,2
ℎ1,𝑀 𝑧𝑡 ~ 𝑁𝐶(0, Σ)
Σ = 𝜎2 𝐼 ~ 𝑎𝑠𝑠𝑢𝑚𝑒𝑑
𝜎2 ~ 𝑘𝑛𝑜𝑤𝑛
ℎ 𝑡|𝛿 ~ 𝑁𝐶(0, 𝐶 𝛿)
𝛿 ~ 𝑝(𝛿)
𝑦𝑡|ℎ 𝑡 ~ 𝑁𝐶(ℎ 𝑡, Σ)
𝑇ℎ𝑒 𝑔𝑜𝑎𝑙 𝑖𝑠 𝑡𝑜 𝑒𝑠𝑡𝑖𝑚𝑎𝑡𝑒 𝒉 𝒕,
𝑓𝑟𝑜𝑚 𝑟𝑒𝑐𝑒𝑖𝑣𝑒𝑑 𝑠𝑖𝑔𝑛𝑎𝑙, 𝑦𝑡
𝑎𝑛𝑑 𝑡ℎ𝑒 𝑑𝑖𝑠𝑡𝑟𝑖𝑏𝑢𝑡𝑖𝑜𝑛𝑠 𝑜𝑓ℎ 𝑡, 𝑦𝑡 𝑎𝑛𝑑 𝑧𝑡
System Model
…
Base station
(M antennas)
UE
(single antenna)
…
ℎ1,1
ℎ1,2
ℎ1,𝑀
MMSE channel estimation
𝑊𝛿 = 𝐶 𝛿(𝐶 𝛿 + 𝛴)−1
𝐶 =
1
𝜎2
𝑡=1
𝑇
𝑦𝑡 𝑦𝑡
𝐻 𝑊∗ 𝐶 =
𝑒𝑥𝑝(𝑡𝑟 𝑊𝛿 𝐶 + 𝑇𝑙𝑜𝑔 𝐼 − 𝑊𝛿 ) 𝑊𝛿 𝑝(𝛿) 𝑑𝛿
𝑒𝑥𝑝(𝑡𝑟 𝑊𝛿 𝐶 + 𝑇𝑙𝑜𝑔 𝐼 − 𝑊𝛿 ) 𝑝(𝛿) 𝑑𝛿
𝐶 𝛿 = 𝑐𝑜𝑣(ℎ 𝑡)
𝑝(𝛿)Distribution of channel parameters,
𝑦𝑡
𝜎2Know noise variance,
Received signal,
Channel covariance matrix
Sample covariance matrix
(estimated)
MMSE filter coefficients
MMSE filter (estimated)
ℎ = 𝑊∗ 𝐶 𝑦
Estimated channel
ℎ 𝑡Generated channel,
General MMSE Estimator:
𝐸 ℎ 𝑡 𝑌, 𝛿 = 𝑊𝛿 𝑦𝑡
ℎ 𝑡 = 𝐸 ℎ 𝑡 𝑌 = 𝑊∗ 𝐶 𝑦𝑡
MMSE channel estimation
𝐶 =
1
𝜎2
𝑡=1
𝑇
𝑦𝑡 𝑦𝑡
𝐻 𝑊∗ 𝐶 =
𝑒𝑥𝑝(𝑡𝑟 𝑊𝛿 𝐶 + 𝑇𝑙𝑜𝑔 𝐼 − 𝑊𝛿 ) 𝑊𝛿 𝑝(𝛿) 𝑑𝛿
𝑒𝑥𝑝(𝑡𝑟 𝑊𝛿 𝐶 + 𝑇𝑙𝑜𝑔 𝐼 − 𝑊𝛿 ) 𝑝(𝛿) 𝑑𝛿
𝑝(𝛿)Distribution of channel parameters,
Sample covariance matrix
(estimated)
MMSE filter (estimated)
ℎ = 𝑾 𝜹 𝑦
Estimated channel
Genie Aided MMSE Estimator (for a lower bound performance):
𝑦𝑡
𝜎2Know noise variance,
Received signal,
𝑾 𝜹 = 𝐶 𝛿(𝐶 𝛿 + 𝛴)−1𝐶 𝛿 = 𝑐𝑜𝑣(ℎ 𝑡)
Channel covariance matrix MMSE filter coefficients
ℎ 𝑡Generated channel,
MMSE channel estimation
𝐶 =
1
𝜎2
𝑡=1
𝑇
𝑦𝑡 𝑦𝑡
𝐻 𝑾 𝑮𝑬 𝑪 =
𝑒𝑥𝑝(𝑡𝑟 𝑊𝛿 𝐶 + 𝑇𝑙𝑜𝑔 𝐼 − 𝑊𝛿 ) 𝑊𝛿 𝑝(𝛿) 𝑑𝛿
𝑒𝑥𝑝(𝑡𝑟 𝑊𝛿 𝐶 + 𝑇𝑙𝑜𝑔 𝐼 − 𝑊𝛿 ) 𝑝(𝛿) 𝑑𝛿𝑦𝑡
𝜎2Know noise variance,
Received signal,
Sample covariance matrix
(estimated)
MMSE filter (estimated)
ℎ = 𝑾 𝑮𝑬 𝑪 𝑦
Estimated channel
Gridded MMSE Estimator:
𝑊𝛿 = 𝐶 𝛿(𝐶 𝛿 + 𝛴)−1𝐶 𝛿 = 𝑐𝑜𝑣(ℎ 𝑡)
Channel covariance matrix MMSE filter coefficients
ℎ 𝑡Generated channel,
Assumption 1: 𝑝 𝛿𝑖 =
1
𝑁
, ∀𝑖 = 1, … , 𝑁
N = 16M
M = number of antennas
MMSE channel estimation
𝐶 =
1
𝜎2
𝑡=1
𝑇
𝑦𝑡 𝑦𝑡
𝐻 𝑾 𝑮𝑬 𝑪 =
𝑒𝑥𝑝(𝑡𝑟 𝑊𝛿 𝐶 + 𝑇𝑙𝑜𝑔 𝐼 − 𝑊𝛿 ) 𝑊𝛿 𝑝(𝛿) 𝑑𝛿
𝑒𝑥𝑝(𝑡𝑟 𝑊𝛿 𝐶 + 𝑇𝑙𝑜𝑔 𝐼 − 𝑊𝛿 ) 𝑝(𝛿) 𝑑𝛿
Assumption 1: 𝑝 𝛿𝑖 =
1
𝑁
, ∀𝑖 = 1, … , 𝑁
N = 16M
M = number of antennas
𝑦𝑡
𝜎2Know noise variance,
Received signal,
Sample covariance matrix
(estimated)
MMSE filter (estimated)
ℎ = 𝑾 𝑮𝑬 𝑪 𝑦
Estimated channel
Gridded MMSE Estimator :
𝑣𝑒𝑐 𝑊𝛿 ∈ 𝐂 𝑀2 𝑥𝑁
→ 𝛰(𝑀2
𝑁) → 𝑂(𝑀3
)
𝑊𝛿 = 𝐶 𝛿(𝐶 𝛿 + 𝛴)−1𝐶 𝛿 = 𝑐𝑜𝑣(ℎ 𝑡)
Channel covariance matrix MMSE filter coefficients
ℎ 𝑡Generated channel,
MMSE channel estimation
𝑪 =
𝟏
𝝈 𝟐
𝒕=𝟏
𝑻
𝑸𝒚 𝒕
𝟐 𝑾 𝑺𝑬 𝑪 =
𝑒𝑥𝑝(𝑡𝑟 𝑊𝑆𝐸 𝐶 + 𝑇𝑙𝑜𝑔 𝐼 − 𝑄 𝐻 𝑊𝛿 𝑄 ) 𝑊𝑆𝐸 𝑝(𝛿) 𝑑𝛿
𝑒𝑥𝑝(𝑡𝑟 𝑊𝑆𝐸 𝐶 + 𝑇𝑙𝑜𝑔 𝐼 − 𝑄 𝐻 𝑊𝛿 𝑄 ) 𝑝(𝛿) 𝑑𝛿
𝑝(𝛿)Distribution of channel parameters,
𝑦𝑡
𝜎2Know noise variance,
Received signal,
Sample covariance matrix
(estimated)
MMSE filter (estimated)
ℎ = 𝑄 𝐻 𝑾 𝑺𝑬 𝑪 𝑄𝑦
Estimated channel
Structured MMSE Estimator:
Assumption 2: 𝑊𝛿 = 𝑄 𝐻
𝑑𝑖𝑎𝑔 𝑤 𝛿 𝑄
𝑤 𝛿 = 𝑑𝑖𝑎𝑔(𝑄𝑊𝛿 𝑄 𝐻
)
𝑄 =
𝐹, 𝑓𝑜𝑟 𝑐𝑖𝑟𝑐𝑢𝑙𝑎𝑛𝑡 𝑒𝑠𝑡𝑖𝑚𝑎𝑡𝑜𝑟
𝐹2, 𝑓𝑜𝑟 𝑡𝑜𝑒𝑝𝑙𝑖𝑡𝑧 𝑒𝑠𝑡𝑖𝑚𝑎𝑡𝑜𝑟
𝑊𝑆𝐸 = 𝑄𝑊𝛿 𝑄 𝐻𝑊𝛿 = 𝐶 𝛿(𝐶 𝛿 + 𝛴)−1𝐶 𝛿 = 𝑐𝑜𝑣(ℎ 𝑡)
Channel covariance matrix MMSE filter coefficients
ℎ 𝑡Generated channel,
MMSE channel estimation
𝑪 =
𝟏
𝝈 𝟐
𝒕=𝟏
𝑻
𝑸𝒚 𝒕
𝟐 𝑾 𝑺𝑬 𝑪 =
𝑒𝑥𝑝(𝑡𝑟 𝑊𝑆𝐸 𝐶 + 𝑇𝑙𝑜𝑔 𝐼 − 𝑄 𝐻 𝑊𝛿 𝑄 ) 𝑊𝑆𝐸 𝑝(𝛿) 𝑑𝛿
𝑒𝑥𝑝(𝑡𝑟 𝑊𝑆𝐸 𝐶 + 𝑇𝑙𝑜𝑔 𝐼 − 𝑄 𝐻 𝑊𝛿 𝑄 ) 𝑝(𝛿) 𝑑𝛿
𝑝(𝛿)Distribution of channel parameters,
𝑦𝑡
𝜎2Know noise variance,
Received signal,
Sample covariance matrix
(estimated)
MMSE filter (estimated)
ℎ = 𝑄 𝐻 𝑾 𝑺𝑬 𝑪 𝑄𝑦
Estimated channel
Structured MMSE Estimator:
𝑊𝑆𝐸 = 𝑄𝑊𝛿 𝑄 𝐻
𝑊𝑆𝐸 ∈ 𝐂 𝑀𝑥𝑁
→ 𝛰(𝑀𝑁) → 𝑂(𝑀2
)
𝑊𝛿 = 𝐶 𝛿(𝐶 𝛿 + 𝛴)−1𝐶 𝛿 = 𝑐𝑜𝑣(ℎ 𝑡)
Channel covariance matrix MMSE filter coefficients
ℎ 𝑡Generated channel,
Assumption 2: 𝑊𝛿 = 𝑄 𝐻
𝑑𝑖𝑎𝑔 𝑤 𝛿 𝑄
𝑤 𝛿 = 𝑑𝑖𝑎𝑔(𝑄𝑊𝛿 𝑄 𝐻
)
𝑄 =
𝐹, 𝑓𝑜𝑟 𝑐𝑖𝑟𝑐𝑢𝑙𝑎𝑛𝑡 𝑒𝑠𝑡𝑖𝑚𝑎𝑡𝑜𝑟
𝐹2, 𝑓𝑜𝑟 𝑡𝑜𝑒𝑝𝑙𝑖𝑡𝑧 𝑒𝑠𝑡𝑖𝑚𝑎𝑡𝑜𝑟
MMSE channel estimation
𝑾 𝑭𝑬 𝑪 =
𝑒𝑥𝑝(𝑡𝑟 𝐹 𝐻 𝑾 𝟎 𝐹 𝐶 + 𝑇𝑙𝑜𝑔 𝐼 − 𝑾 𝑭𝑬 ) 𝐹 𝐻 𝑾 𝟎 𝐹𝑝(𝛿) 𝑑𝛿
𝑒𝑥𝑝(𝑡𝑟 𝐹 𝐻 𝑾 𝟎 𝐹 𝐶 + 𝑇𝑙𝑜𝑔 𝐼 − 𝑾 𝑭𝑬 ) 𝑝(𝛿) 𝑑𝛿
𝑝(𝛿)Distribution of channel parameters,
𝑦𝑡
𝜎2Know noise variance,
Received signal,
Sample covariance matrix
(estimated)
MMSE filter (estimated)
ℎ = 𝐹 𝐻 𝑾 𝑭𝑬 𝑪 𝐹𝑦
Estimated channel
Fast MMSE Estimator:
Generated channel, 𝑾 𝑭𝑬 = 𝐶𝑐(𝐶𝑐 + 𝛴)−1𝐶 𝛿 = 𝑐𝑜𝑣(ℎ 𝑡)
Channel covariance matrix MMSE filter coefficients
ℎ 𝑡Generated channel, 𝑪 𝒄 = 𝑐𝑖𝑟𝑐(𝐶 𝛿)
Circulant approximation of
channel covariance matrix
𝑾 𝟎 = 𝐹𝑾 𝑭𝑬
FFT of filter
𝑪 =
𝟏
𝝈 𝟐
𝒕=𝟏
𝑻
𝐹𝒚 𝒕
𝟐
Assumption 3: 𝐯𝐞𝐜(𝑊𝛿) = 𝐹 𝐻
𝑑𝑖𝑎𝑔 𝐹𝑤 𝛿 𝐹
MMSE channel estimation
𝑾 𝑭𝑬 𝑪 =
𝑒𝑥𝑝(𝑡𝑟 𝐹 𝐻 𝑾 𝟎 𝐹 𝐶 + 𝑇𝑙𝑜𝑔 𝐼 − 𝑾 𝑭𝑬 ) 𝐹 𝐻 𝑾 𝟎 𝐹𝑝(𝛿) 𝑑𝛿
𝑒𝑥𝑝(𝑡𝑟 𝐹 𝐻 𝑾 𝟎 𝐹 𝐶 + 𝑇𝑙𝑜𝑔 𝐼 − 𝑾 𝑭𝑬 ) 𝑝(𝛿) 𝑑𝛿
𝑝(𝛿)Distribution of channel parameters,
𝑦𝑡
𝜎2Know noise variance,
Received signal,
Sample covariance matrix
(estimated)
MMSE filter (estimated)
ℎ = 𝐹 𝐻 𝑾 𝑭𝑬 𝑪 𝐹𝑦
Estimated channel
Fast MMSE Estimator:
Generated channel, 𝑾 𝑭𝑬 = 𝐶𝑐(𝐶𝑐 + 𝛴)−1𝐶 𝛿 = 𝑐𝑜𝑣(ℎ 𝑡)
Channel covariance matrix MMSE filter coefficients
ℎ 𝑡Generated channel, 𝑪 𝒄 = 𝑐𝑖𝑟𝑐(𝐶 𝛿)
Circulant approximation of
channel covariance matrix
𝑾 𝟎 = 𝐹𝑾 𝑭𝑬
FFT of filter
𝑪 =
𝟏
𝝈 𝟐
𝒕=𝟏
𝑻
𝐹𝒚 𝒕
𝟐
𝑊𝑆𝐸 𝑐𝑜𝑚𝑝𝑢𝑡𝑎𝑡𝑖𝑜𝑛 → 𝛰(𝑀log𝑀)
MMSE channel estimation
𝑾 𝑭𝑬 𝑪 =
𝑒𝑥𝑝(𝑡𝑟 𝐹 𝐻 𝑾 𝟎 𝐹 𝐶 + 𝑇𝑙𝑜𝑔 𝐼 − 𝑾 𝑭𝑬 ) 𝐹 𝐻 𝑾 𝟎 𝐹𝑝(𝛿) 𝑑𝛿
𝑒𝑥𝑝(𝑡𝑟 𝐹 𝐻 𝑾 𝟎 𝐹 𝐶 + 𝑇𝑙𝑜𝑔 𝐼 − 𝑾 𝑭𝑬 ) 𝑝(𝛿) 𝑑𝛿
𝑝(𝛿)Distribution of channel parameters,
𝑦𝑡
𝜎2Know noise variance,
Received signal,
Sample covariance matrix
(estimated)
MMSE filter (estimated)
ℎ = 𝐹 𝐻 𝑾 𝑭𝑬 𝑪 𝐹𝑦
Estimated channel
Motivation for Low-Complexity Neural Network:
Generated channel, 𝑾 𝑭𝑬 = 𝐶𝑐(𝐶𝑐 + 𝛴)−1𝐶 𝛿 = 𝑐𝑜𝑣(ℎ 𝑡)
Channel covariance matrix MMSE filter coefficients
ℎ 𝑡Generated channel, 𝑪 𝒄 = 𝑐𝑖𝑟𝑐(𝐶 𝛿)
Circulant approximation of
channel covariance matrix
𝑾 𝟎 = 𝐹𝑾 𝑭𝑬
FFT of filter
𝑪 =
𝟏
𝝈 𝟐
𝒕=𝟏
𝑻
𝐹𝒚 𝒕
𝟐
𝑭 𝑯
𝒅𝒊𝒂𝒈 𝑭𝒂 𝑭𝒙 = 𝒂 ∗ 𝒙
MMSE channel estimation
Learning Fast MMSE Estimator using a Convolutional Neural Network:
𝑦𝑡
𝜎2Know noise variance,
Received signal,
ℎ = 𝐹 𝐻 𝑾 𝑪𝑵𝑵 𝐹𝑦
Estimated channel
ℎ 𝑡Generated channel,
Conv ReLU Conv
Bias Bias
𝑾 𝑪𝑵𝑵
𝑦𝑡Generated channel,
Parameter Value
Epochs 10 000
Training batch size 600
Testing batch size 100
Learning rate (64/M)*1e-4
Convolutional layer 2
Hierarchical Training
1. Train CNN for M antennas
2. Interpolate kernel from length M to 2M
3. Use interpolated values of kernel from previous step
for initializing CNN when training for M antennas.
MMSE channel estimation
Learning Fast MMSE Estimator using a Convolutional Neural Network:
ℎ 𝑡Generated channel,
Conv ReLU Conv
Bias Bias
𝑾 𝑪𝑵𝑵
𝑦𝑡Generated channel,
MSE per antenna at an SNR of 0 dB for estimation from a single snapshot (T=1). Channel model with
one propagation path with uniformly distributed angle.
MSE per antenna at an SNR of 0 dB for estimation from a single snapshot (T=1). Channel model with
three propagation paths.
MSE per antenna for M = 64 antennas and for estimation from a single snapshot (T = 1). Channel
model with three propagation paths.
Thank You

More Related Content

Similar to Learning the mmse channel estimators

Blind PNLMS Adaptive Algorithm for SIMO FIR Channel Estimation
Blind PNLMS  Adaptive Algorithm for SIMO FIR Channel EstimationBlind PNLMS  Adaptive Algorithm for SIMO FIR Channel Estimation
Blind PNLMS Adaptive Algorithm for SIMO FIR Channel Estimationardodul
 
Adaptive analog beamforming
Adaptive analog beamformingAdaptive analog beamforming
Adaptive analog beamformingKhalid Hussain
 
Communication Systems_B.P. Lathi and Zhi Ding (Lecture No 22-30)
Communication Systems_B.P. Lathi and Zhi Ding (Lecture No 22-30)Communication Systems_B.P. Lathi and Zhi Ding (Lecture No 22-30)
Communication Systems_B.P. Lathi and Zhi Ding (Lecture No 22-30)Adnan Zafar
 
Performance Comparison of Modified Variable Step Size Leaky LMS Algorithm for...
Performance Comparison of Modified Variable Step Size Leaky LMS Algorithm for...Performance Comparison of Modified Variable Step Size Leaky LMS Algorithm for...
Performance Comparison of Modified Variable Step Size Leaky LMS Algorithm for...ijcnac
 
Project session part_I
Project  session part_IProject  session part_I
Project session part_IMina Yonan
 
Channel Equalisation
Channel EqualisationChannel Equalisation
Channel EqualisationPoonan Sahoo
 
Bandpass Differentiator FM Receiver
Bandpass Differentiator FM ReceiverBandpass Differentiator FM Receiver
Bandpass Differentiator FM ReceiverNathan Wendt
 
Channel Equalization of WCDMA Downlink System Using Finite Length MMSE-DFE
Channel Equalization of WCDMA Downlink System Using Finite Length MMSE-DFEChannel Equalization of WCDMA Downlink System Using Finite Length MMSE-DFE
Channel Equalization of WCDMA Downlink System Using Finite Length MMSE-DFEIOSR Journals
 
Channel Equalization of WCDMA Downlink System Using Finite Length MMSE-DFE
Channel Equalization of WCDMA Downlink System Using Finite Length MMSE-DFEChannel Equalization of WCDMA Downlink System Using Finite Length MMSE-DFE
Channel Equalization of WCDMA Downlink System Using Finite Length MMSE-DFEIOSR Journals
 
Voice Activity Detection using Single Frequency Filtering
Voice Activity Detection using Single Frequency FilteringVoice Activity Detection using Single Frequency Filtering
Voice Activity Detection using Single Frequency FilteringTejus Adiga M
 
Equalization
EqualizationEqualization
Equalizationbhabendu
 
Support Vector Machine Techniques for Nonlinear Equalization
Support Vector Machine Techniques for Nonlinear EqualizationSupport Vector Machine Techniques for Nonlinear Equalization
Support Vector Machine Techniques for Nonlinear EqualizationShamman Noor Shoudha
 
Intelligent Selection of Transmission Waveform for Active Sonar Using Predict...
Intelligent Selection of Transmission Waveform for Active Sonar Using Predict...Intelligent Selection of Transmission Waveform for Active Sonar Using Predict...
Intelligent Selection of Transmission Waveform for Active Sonar Using Predict...Venkata Sasikiran Veeramachaneni
 
Fundamentals of music processing chapter 5 발표자료
Fundamentals of music processing chapter 5 발표자료Fundamentals of music processing chapter 5 발표자료
Fundamentals of music processing chapter 5 발표자료Jeong Choi
 

Similar to Learning the mmse channel estimators (20)

Icmmse slides
Icmmse slidesIcmmse slides
Icmmse slides
 
Blind PNLMS Adaptive Algorithm for SIMO FIR Channel Estimation
Blind PNLMS  Adaptive Algorithm for SIMO FIR Channel EstimationBlind PNLMS  Adaptive Algorithm for SIMO FIR Channel Estimation
Blind PNLMS Adaptive Algorithm for SIMO FIR Channel Estimation
 
Adaptive analog beamforming
Adaptive analog beamformingAdaptive analog beamforming
Adaptive analog beamforming
 
Communication Systems_B.P. Lathi and Zhi Ding (Lecture No 22-30)
Communication Systems_B.P. Lathi and Zhi Ding (Lecture No 22-30)Communication Systems_B.P. Lathi and Zhi Ding (Lecture No 22-30)
Communication Systems_B.P. Lathi and Zhi Ding (Lecture No 22-30)
 
Amplitude modulated-systems
Amplitude modulated-systemsAmplitude modulated-systems
Amplitude modulated-systems
 
Performance Comparison of Modified Variable Step Size Leaky LMS Algorithm for...
Performance Comparison of Modified Variable Step Size Leaky LMS Algorithm for...Performance Comparison of Modified Variable Step Size Leaky LMS Algorithm for...
Performance Comparison of Modified Variable Step Size Leaky LMS Algorithm for...
 
Project session part_I
Project  session part_IProject  session part_I
Project session part_I
 
ADC PPT.pptx
ADC PPT.pptxADC PPT.pptx
ADC PPT.pptx
 
Channel Equalisation
Channel EqualisationChannel Equalisation
Channel Equalisation
 
Bandpass Differentiator FM Receiver
Bandpass Differentiator FM ReceiverBandpass Differentiator FM Receiver
Bandpass Differentiator FM Receiver
 
Channel Equalization of WCDMA Downlink System Using Finite Length MMSE-DFE
Channel Equalization of WCDMA Downlink System Using Finite Length MMSE-DFEChannel Equalization of WCDMA Downlink System Using Finite Length MMSE-DFE
Channel Equalization of WCDMA Downlink System Using Finite Length MMSE-DFE
 
Channel Equalization of WCDMA Downlink System Using Finite Length MMSE-DFE
Channel Equalization of WCDMA Downlink System Using Finite Length MMSE-DFEChannel Equalization of WCDMA Downlink System Using Finite Length MMSE-DFE
Channel Equalization of WCDMA Downlink System Using Finite Length MMSE-DFE
 
Voice Activity Detection using Single Frequency Filtering
Voice Activity Detection using Single Frequency FilteringVoice Activity Detection using Single Frequency Filtering
Voice Activity Detection using Single Frequency Filtering
 
Channel equalization
Channel equalizationChannel equalization
Channel equalization
 
Equalization
EqualizationEqualization
Equalization
 
fading channels
 fading channels fading channels
fading channels
 
Support Vector Machine Techniques for Nonlinear Equalization
Support Vector Machine Techniques for Nonlinear EqualizationSupport Vector Machine Techniques for Nonlinear Equalization
Support Vector Machine Techniques for Nonlinear Equalization
 
Intelligent Selection of Transmission Waveform for Active Sonar Using Predict...
Intelligent Selection of Transmission Waveform for Active Sonar Using Predict...Intelligent Selection of Transmission Waveform for Active Sonar Using Predict...
Intelligent Selection of Transmission Waveform for Active Sonar Using Predict...
 
Fundamentals of music processing chapter 5 발표자료
Fundamentals of music processing chapter 5 발표자료Fundamentals of music processing chapter 5 발표자료
Fundamentals of music processing chapter 5 발표자료
 
Amplitude modulated-systmes
Amplitude modulated-systmesAmplitude modulated-systmes
Amplitude modulated-systmes
 

Recently uploaded

FULL ENJOY Call Girls In Mahipalpur Delhi Contact Us 8377877756
FULL ENJOY Call Girls In Mahipalpur Delhi Contact Us 8377877756FULL ENJOY Call Girls In Mahipalpur Delhi Contact Us 8377877756
FULL ENJOY Call Girls In Mahipalpur Delhi Contact Us 8377877756dollysharma2066
 
Unleashing the Power of the SORA AI lastest leap
Unleashing the Power of the SORA AI lastest leapUnleashing the Power of the SORA AI lastest leap
Unleashing the Power of the SORA AI lastest leapRishantSharmaFr
 
Block diagram reduction techniques in control systems.ppt
Block diagram reduction techniques in control systems.pptBlock diagram reduction techniques in control systems.ppt
Block diagram reduction techniques in control systems.pptNANDHAKUMARA10
 
Top Rated Call Girls In chittoor 📱 {7001035870} VIP Escorts chittoor
Top Rated Call Girls In chittoor 📱 {7001035870} VIP Escorts chittoorTop Rated Call Girls In chittoor 📱 {7001035870} VIP Escorts chittoor
Top Rated Call Girls In chittoor 📱 {7001035870} VIP Escorts chittoordharasingh5698
 
VIP Call Girls Ankleshwar 7001035870 Whatsapp Number, 24/07 Booking
VIP Call Girls Ankleshwar 7001035870 Whatsapp Number, 24/07 BookingVIP Call Girls Ankleshwar 7001035870 Whatsapp Number, 24/07 Booking
VIP Call Girls Ankleshwar 7001035870 Whatsapp Number, 24/07 Bookingdharasingh5698
 
Call Girls In Bangalore ☎ 7737669865 🥵 Book Your One night Stand
Call Girls In Bangalore ☎ 7737669865 🥵 Book Your One night StandCall Girls In Bangalore ☎ 7737669865 🥵 Book Your One night Stand
Call Girls In Bangalore ☎ 7737669865 🥵 Book Your One night Standamitlee9823
 
Design For Accessibility: Getting it right from the start
Design For Accessibility: Getting it right from the startDesign For Accessibility: Getting it right from the start
Design For Accessibility: Getting it right from the startQuintin Balsdon
 
Bhosari ( Call Girls ) Pune 6297143586 Hot Model With Sexy Bhabi Ready For ...
Bhosari ( Call Girls ) Pune  6297143586  Hot Model With Sexy Bhabi Ready For ...Bhosari ( Call Girls ) Pune  6297143586  Hot Model With Sexy Bhabi Ready For ...
Bhosari ( Call Girls ) Pune 6297143586 Hot Model With Sexy Bhabi Ready For ...tanu pandey
 
Thermal Engineering -unit - III & IV.ppt
Thermal Engineering -unit - III & IV.pptThermal Engineering -unit - III & IV.ppt
Thermal Engineering -unit - III & IV.pptDineshKumar4165
 
KubeKraft presentation @CloudNativeHooghly
KubeKraft presentation @CloudNativeHooghlyKubeKraft presentation @CloudNativeHooghly
KubeKraft presentation @CloudNativeHooghlysanyuktamishra911
 
Hostel management system project report..pdf
Hostel management system project report..pdfHostel management system project report..pdf
Hostel management system project report..pdfKamal Acharya
 
Generative AI or GenAI technology based PPT
Generative AI or GenAI technology based PPTGenerative AI or GenAI technology based PPT
Generative AI or GenAI technology based PPTbhaskargani46
 
Employee leave management system project.
Employee leave management system project.Employee leave management system project.
Employee leave management system project.Kamal Acharya
 
Work-Permit-Receiver-in-Saudi-Aramco.pptx
Work-Permit-Receiver-in-Saudi-Aramco.pptxWork-Permit-Receiver-in-Saudi-Aramco.pptx
Work-Permit-Receiver-in-Saudi-Aramco.pptxJuliansyahHarahap1
 
chapter 5.pptx: drainage and irrigation engineering
chapter 5.pptx: drainage and irrigation engineeringchapter 5.pptx: drainage and irrigation engineering
chapter 5.pptx: drainage and irrigation engineeringmulugeta48
 

Recently uploaded (20)

FULL ENJOY Call Girls In Mahipalpur Delhi Contact Us 8377877756
FULL ENJOY Call Girls In Mahipalpur Delhi Contact Us 8377877756FULL ENJOY Call Girls In Mahipalpur Delhi Contact Us 8377877756
FULL ENJOY Call Girls In Mahipalpur Delhi Contact Us 8377877756
 
Unleashing the Power of the SORA AI lastest leap
Unleashing the Power of the SORA AI lastest leapUnleashing the Power of the SORA AI lastest leap
Unleashing the Power of the SORA AI lastest leap
 
Block diagram reduction techniques in control systems.ppt
Block diagram reduction techniques in control systems.pptBlock diagram reduction techniques in control systems.ppt
Block diagram reduction techniques in control systems.ppt
 
Top Rated Call Girls In chittoor 📱 {7001035870} VIP Escorts chittoor
Top Rated Call Girls In chittoor 📱 {7001035870} VIP Escorts chittoorTop Rated Call Girls In chittoor 📱 {7001035870} VIP Escorts chittoor
Top Rated Call Girls In chittoor 📱 {7001035870} VIP Escorts chittoor
 
Water Industry Process Automation & Control Monthly - April 2024
Water Industry Process Automation & Control Monthly - April 2024Water Industry Process Automation & Control Monthly - April 2024
Water Industry Process Automation & Control Monthly - April 2024
 
VIP Call Girls Ankleshwar 7001035870 Whatsapp Number, 24/07 Booking
VIP Call Girls Ankleshwar 7001035870 Whatsapp Number, 24/07 BookingVIP Call Girls Ankleshwar 7001035870 Whatsapp Number, 24/07 Booking
VIP Call Girls Ankleshwar 7001035870 Whatsapp Number, 24/07 Booking
 
Call Girls In Bangalore ☎ 7737669865 🥵 Book Your One night Stand
Call Girls In Bangalore ☎ 7737669865 🥵 Book Your One night StandCall Girls In Bangalore ☎ 7737669865 🥵 Book Your One night Stand
Call Girls In Bangalore ☎ 7737669865 🥵 Book Your One night Stand
 
Design For Accessibility: Getting it right from the start
Design For Accessibility: Getting it right from the startDesign For Accessibility: Getting it right from the start
Design For Accessibility: Getting it right from the start
 
Call Now ≽ 9953056974 ≼🔝 Call Girls In New Ashok Nagar ≼🔝 Delhi door step de...
Call Now ≽ 9953056974 ≼🔝 Call Girls In New Ashok Nagar  ≼🔝 Delhi door step de...Call Now ≽ 9953056974 ≼🔝 Call Girls In New Ashok Nagar  ≼🔝 Delhi door step de...
Call Now ≽ 9953056974 ≼🔝 Call Girls In New Ashok Nagar ≼🔝 Delhi door step de...
 
Bhosari ( Call Girls ) Pune 6297143586 Hot Model With Sexy Bhabi Ready For ...
Bhosari ( Call Girls ) Pune  6297143586  Hot Model With Sexy Bhabi Ready For ...Bhosari ( Call Girls ) Pune  6297143586  Hot Model With Sexy Bhabi Ready For ...
Bhosari ( Call Girls ) Pune 6297143586 Hot Model With Sexy Bhabi Ready For ...
 
(INDIRA) Call Girl Aurangabad Call Now 8617697112 Aurangabad Escorts 24x7
(INDIRA) Call Girl Aurangabad Call Now 8617697112 Aurangabad Escorts 24x7(INDIRA) Call Girl Aurangabad Call Now 8617697112 Aurangabad Escorts 24x7
(INDIRA) Call Girl Aurangabad Call Now 8617697112 Aurangabad Escorts 24x7
 
Thermal Engineering -unit - III & IV.ppt
Thermal Engineering -unit - III & IV.pptThermal Engineering -unit - III & IV.ppt
Thermal Engineering -unit - III & IV.ppt
 
Call Girls in Netaji Nagar, Delhi 💯 Call Us 🔝9953056974 🔝 Escort Service
Call Girls in Netaji Nagar, Delhi 💯 Call Us 🔝9953056974 🔝 Escort ServiceCall Girls in Netaji Nagar, Delhi 💯 Call Us 🔝9953056974 🔝 Escort Service
Call Girls in Netaji Nagar, Delhi 💯 Call Us 🔝9953056974 🔝 Escort Service
 
KubeKraft presentation @CloudNativeHooghly
KubeKraft presentation @CloudNativeHooghlyKubeKraft presentation @CloudNativeHooghly
KubeKraft presentation @CloudNativeHooghly
 
Hostel management system project report..pdf
Hostel management system project report..pdfHostel management system project report..pdf
Hostel management system project report..pdf
 
Generative AI or GenAI technology based PPT
Generative AI or GenAI technology based PPTGenerative AI or GenAI technology based PPT
Generative AI or GenAI technology based PPT
 
Employee leave management system project.
Employee leave management system project.Employee leave management system project.
Employee leave management system project.
 
Work-Permit-Receiver-in-Saudi-Aramco.pptx
Work-Permit-Receiver-in-Saudi-Aramco.pptxWork-Permit-Receiver-in-Saudi-Aramco.pptx
Work-Permit-Receiver-in-Saudi-Aramco.pptx
 
chapter 5.pptx: drainage and irrigation engineering
chapter 5.pptx: drainage and irrigation engineeringchapter 5.pptx: drainage and irrigation engineering
chapter 5.pptx: drainage and irrigation engineering
 
(INDIRA) Call Girl Meerut Call Now 8617697112 Meerut Escorts 24x7
(INDIRA) Call Girl Meerut Call Now 8617697112 Meerut Escorts 24x7(INDIRA) Call Girl Meerut Call Now 8617697112 Meerut Escorts 24x7
(INDIRA) Call Girl Meerut Call Now 8617697112 Meerut Escorts 24x7
 

Learning the mmse channel estimators

  • 1. Learning the MMSE Channel Estimator EESC 6353 Project Shamman Noor
  • 2. Introduction • System Model • MMSE Estimation: - Discrete MMSE estimation - Structured MMSE estimation - Circulant structure - Toeplitz structure - Fast MMSE estimation - CNN based fast MMSE estimation • CNN Model and parameters Software used: • MATLAB – 2018a • Python – 3.7 • TensorFlow – 1.13.1
  • 3. System Model … Base station (M antennas) UE (single antenna) … 𝑦𝑡 = ℎ 𝑡 + 𝑧𝑡 ℎ1,1 ℎ1,2 ℎ1,𝑀 𝑧𝑡 ~ 𝑁𝐶(0, Σ) Σ = 𝜎2 𝐼 ~ 𝑎𝑠𝑠𝑢𝑚𝑒𝑑 𝜎2 ~ 𝑘𝑛𝑜𝑤𝑛 ℎ 𝑡|𝛿 ~ 𝑁𝐶(0, 𝐶 𝛿) 𝛿 ~ 𝑝(𝛿) 𝑦𝑡|ℎ 𝑡 ~ 𝑁𝐶(ℎ 𝑡, Σ) 𝑇ℎ𝑒 𝑔𝑜𝑎𝑙 𝑖𝑠 𝑡𝑜 𝑒𝑠𝑡𝑖𝑚𝑎𝑡𝑒 𝒉 𝒕, 𝑓𝑟𝑜𝑚 𝑟𝑒𝑐𝑒𝑖𝑣𝑒𝑑 𝑠𝑖𝑔𝑛𝑎𝑙, 𝑦𝑡 𝑎𝑛𝑑 𝑡ℎ𝑒 𝑑𝑖𝑠𝑡𝑟𝑖𝑏𝑢𝑡𝑖𝑜𝑛𝑠 𝑜𝑓ℎ 𝑡, 𝑦𝑡 𝑎𝑛𝑑 𝑧𝑡
  • 4. System Model … Base station (M antennas) UE (single antenna) … ℎ1,1 ℎ1,2 ℎ1,𝑀
  • 5. MMSE channel estimation 𝑊𝛿 = 𝐶 𝛿(𝐶 𝛿 + 𝛴)−1 𝐶 = 1 𝜎2 𝑡=1 𝑇 𝑦𝑡 𝑦𝑡 𝐻 𝑊∗ 𝐶 = 𝑒𝑥𝑝(𝑡𝑟 𝑊𝛿 𝐶 + 𝑇𝑙𝑜𝑔 𝐼 − 𝑊𝛿 ) 𝑊𝛿 𝑝(𝛿) 𝑑𝛿 𝑒𝑥𝑝(𝑡𝑟 𝑊𝛿 𝐶 + 𝑇𝑙𝑜𝑔 𝐼 − 𝑊𝛿 ) 𝑝(𝛿) 𝑑𝛿 𝐶 𝛿 = 𝑐𝑜𝑣(ℎ 𝑡) 𝑝(𝛿)Distribution of channel parameters, 𝑦𝑡 𝜎2Know noise variance, Received signal, Channel covariance matrix Sample covariance matrix (estimated) MMSE filter coefficients MMSE filter (estimated) ℎ = 𝑊∗ 𝐶 𝑦 Estimated channel ℎ 𝑡Generated channel, General MMSE Estimator: 𝐸 ℎ 𝑡 𝑌, 𝛿 = 𝑊𝛿 𝑦𝑡 ℎ 𝑡 = 𝐸 ℎ 𝑡 𝑌 = 𝑊∗ 𝐶 𝑦𝑡
  • 6. MMSE channel estimation 𝐶 = 1 𝜎2 𝑡=1 𝑇 𝑦𝑡 𝑦𝑡 𝐻 𝑊∗ 𝐶 = 𝑒𝑥𝑝(𝑡𝑟 𝑊𝛿 𝐶 + 𝑇𝑙𝑜𝑔 𝐼 − 𝑊𝛿 ) 𝑊𝛿 𝑝(𝛿) 𝑑𝛿 𝑒𝑥𝑝(𝑡𝑟 𝑊𝛿 𝐶 + 𝑇𝑙𝑜𝑔 𝐼 − 𝑊𝛿 ) 𝑝(𝛿) 𝑑𝛿 𝑝(𝛿)Distribution of channel parameters, Sample covariance matrix (estimated) MMSE filter (estimated) ℎ = 𝑾 𝜹 𝑦 Estimated channel Genie Aided MMSE Estimator (for a lower bound performance): 𝑦𝑡 𝜎2Know noise variance, Received signal, 𝑾 𝜹 = 𝐶 𝛿(𝐶 𝛿 + 𝛴)−1𝐶 𝛿 = 𝑐𝑜𝑣(ℎ 𝑡) Channel covariance matrix MMSE filter coefficients ℎ 𝑡Generated channel,
  • 7. MMSE channel estimation 𝐶 = 1 𝜎2 𝑡=1 𝑇 𝑦𝑡 𝑦𝑡 𝐻 𝑾 𝑮𝑬 𝑪 = 𝑒𝑥𝑝(𝑡𝑟 𝑊𝛿 𝐶 + 𝑇𝑙𝑜𝑔 𝐼 − 𝑊𝛿 ) 𝑊𝛿 𝑝(𝛿) 𝑑𝛿 𝑒𝑥𝑝(𝑡𝑟 𝑊𝛿 𝐶 + 𝑇𝑙𝑜𝑔 𝐼 − 𝑊𝛿 ) 𝑝(𝛿) 𝑑𝛿𝑦𝑡 𝜎2Know noise variance, Received signal, Sample covariance matrix (estimated) MMSE filter (estimated) ℎ = 𝑾 𝑮𝑬 𝑪 𝑦 Estimated channel Gridded MMSE Estimator: 𝑊𝛿 = 𝐶 𝛿(𝐶 𝛿 + 𝛴)−1𝐶 𝛿 = 𝑐𝑜𝑣(ℎ 𝑡) Channel covariance matrix MMSE filter coefficients ℎ 𝑡Generated channel, Assumption 1: 𝑝 𝛿𝑖 = 1 𝑁 , ∀𝑖 = 1, … , 𝑁 N = 16M M = number of antennas
  • 8. MMSE channel estimation 𝐶 = 1 𝜎2 𝑡=1 𝑇 𝑦𝑡 𝑦𝑡 𝐻 𝑾 𝑮𝑬 𝑪 = 𝑒𝑥𝑝(𝑡𝑟 𝑊𝛿 𝐶 + 𝑇𝑙𝑜𝑔 𝐼 − 𝑊𝛿 ) 𝑊𝛿 𝑝(𝛿) 𝑑𝛿 𝑒𝑥𝑝(𝑡𝑟 𝑊𝛿 𝐶 + 𝑇𝑙𝑜𝑔 𝐼 − 𝑊𝛿 ) 𝑝(𝛿) 𝑑𝛿 Assumption 1: 𝑝 𝛿𝑖 = 1 𝑁 , ∀𝑖 = 1, … , 𝑁 N = 16M M = number of antennas 𝑦𝑡 𝜎2Know noise variance, Received signal, Sample covariance matrix (estimated) MMSE filter (estimated) ℎ = 𝑾 𝑮𝑬 𝑪 𝑦 Estimated channel Gridded MMSE Estimator : 𝑣𝑒𝑐 𝑊𝛿 ∈ 𝐂 𝑀2 𝑥𝑁 → 𝛰(𝑀2 𝑁) → 𝑂(𝑀3 ) 𝑊𝛿 = 𝐶 𝛿(𝐶 𝛿 + 𝛴)−1𝐶 𝛿 = 𝑐𝑜𝑣(ℎ 𝑡) Channel covariance matrix MMSE filter coefficients ℎ 𝑡Generated channel,
  • 9. MMSE channel estimation 𝑪 = 𝟏 𝝈 𝟐 𝒕=𝟏 𝑻 𝑸𝒚 𝒕 𝟐 𝑾 𝑺𝑬 𝑪 = 𝑒𝑥𝑝(𝑡𝑟 𝑊𝑆𝐸 𝐶 + 𝑇𝑙𝑜𝑔 𝐼 − 𝑄 𝐻 𝑊𝛿 𝑄 ) 𝑊𝑆𝐸 𝑝(𝛿) 𝑑𝛿 𝑒𝑥𝑝(𝑡𝑟 𝑊𝑆𝐸 𝐶 + 𝑇𝑙𝑜𝑔 𝐼 − 𝑄 𝐻 𝑊𝛿 𝑄 ) 𝑝(𝛿) 𝑑𝛿 𝑝(𝛿)Distribution of channel parameters, 𝑦𝑡 𝜎2Know noise variance, Received signal, Sample covariance matrix (estimated) MMSE filter (estimated) ℎ = 𝑄 𝐻 𝑾 𝑺𝑬 𝑪 𝑄𝑦 Estimated channel Structured MMSE Estimator: Assumption 2: 𝑊𝛿 = 𝑄 𝐻 𝑑𝑖𝑎𝑔 𝑤 𝛿 𝑄 𝑤 𝛿 = 𝑑𝑖𝑎𝑔(𝑄𝑊𝛿 𝑄 𝐻 ) 𝑄 = 𝐹, 𝑓𝑜𝑟 𝑐𝑖𝑟𝑐𝑢𝑙𝑎𝑛𝑡 𝑒𝑠𝑡𝑖𝑚𝑎𝑡𝑜𝑟 𝐹2, 𝑓𝑜𝑟 𝑡𝑜𝑒𝑝𝑙𝑖𝑡𝑧 𝑒𝑠𝑡𝑖𝑚𝑎𝑡𝑜𝑟 𝑊𝑆𝐸 = 𝑄𝑊𝛿 𝑄 𝐻𝑊𝛿 = 𝐶 𝛿(𝐶 𝛿 + 𝛴)−1𝐶 𝛿 = 𝑐𝑜𝑣(ℎ 𝑡) Channel covariance matrix MMSE filter coefficients ℎ 𝑡Generated channel,
  • 10. MMSE channel estimation 𝑪 = 𝟏 𝝈 𝟐 𝒕=𝟏 𝑻 𝑸𝒚 𝒕 𝟐 𝑾 𝑺𝑬 𝑪 = 𝑒𝑥𝑝(𝑡𝑟 𝑊𝑆𝐸 𝐶 + 𝑇𝑙𝑜𝑔 𝐼 − 𝑄 𝐻 𝑊𝛿 𝑄 ) 𝑊𝑆𝐸 𝑝(𝛿) 𝑑𝛿 𝑒𝑥𝑝(𝑡𝑟 𝑊𝑆𝐸 𝐶 + 𝑇𝑙𝑜𝑔 𝐼 − 𝑄 𝐻 𝑊𝛿 𝑄 ) 𝑝(𝛿) 𝑑𝛿 𝑝(𝛿)Distribution of channel parameters, 𝑦𝑡 𝜎2Know noise variance, Received signal, Sample covariance matrix (estimated) MMSE filter (estimated) ℎ = 𝑄 𝐻 𝑾 𝑺𝑬 𝑪 𝑄𝑦 Estimated channel Structured MMSE Estimator: 𝑊𝑆𝐸 = 𝑄𝑊𝛿 𝑄 𝐻 𝑊𝑆𝐸 ∈ 𝐂 𝑀𝑥𝑁 → 𝛰(𝑀𝑁) → 𝑂(𝑀2 ) 𝑊𝛿 = 𝐶 𝛿(𝐶 𝛿 + 𝛴)−1𝐶 𝛿 = 𝑐𝑜𝑣(ℎ 𝑡) Channel covariance matrix MMSE filter coefficients ℎ 𝑡Generated channel, Assumption 2: 𝑊𝛿 = 𝑄 𝐻 𝑑𝑖𝑎𝑔 𝑤 𝛿 𝑄 𝑤 𝛿 = 𝑑𝑖𝑎𝑔(𝑄𝑊𝛿 𝑄 𝐻 ) 𝑄 = 𝐹, 𝑓𝑜𝑟 𝑐𝑖𝑟𝑐𝑢𝑙𝑎𝑛𝑡 𝑒𝑠𝑡𝑖𝑚𝑎𝑡𝑜𝑟 𝐹2, 𝑓𝑜𝑟 𝑡𝑜𝑒𝑝𝑙𝑖𝑡𝑧 𝑒𝑠𝑡𝑖𝑚𝑎𝑡𝑜𝑟
  • 11. MMSE channel estimation 𝑾 𝑭𝑬 𝑪 = 𝑒𝑥𝑝(𝑡𝑟 𝐹 𝐻 𝑾 𝟎 𝐹 𝐶 + 𝑇𝑙𝑜𝑔 𝐼 − 𝑾 𝑭𝑬 ) 𝐹 𝐻 𝑾 𝟎 𝐹𝑝(𝛿) 𝑑𝛿 𝑒𝑥𝑝(𝑡𝑟 𝐹 𝐻 𝑾 𝟎 𝐹 𝐶 + 𝑇𝑙𝑜𝑔 𝐼 − 𝑾 𝑭𝑬 ) 𝑝(𝛿) 𝑑𝛿 𝑝(𝛿)Distribution of channel parameters, 𝑦𝑡 𝜎2Know noise variance, Received signal, Sample covariance matrix (estimated) MMSE filter (estimated) ℎ = 𝐹 𝐻 𝑾 𝑭𝑬 𝑪 𝐹𝑦 Estimated channel Fast MMSE Estimator: Generated channel, 𝑾 𝑭𝑬 = 𝐶𝑐(𝐶𝑐 + 𝛴)−1𝐶 𝛿 = 𝑐𝑜𝑣(ℎ 𝑡) Channel covariance matrix MMSE filter coefficients ℎ 𝑡Generated channel, 𝑪 𝒄 = 𝑐𝑖𝑟𝑐(𝐶 𝛿) Circulant approximation of channel covariance matrix 𝑾 𝟎 = 𝐹𝑾 𝑭𝑬 FFT of filter 𝑪 = 𝟏 𝝈 𝟐 𝒕=𝟏 𝑻 𝐹𝒚 𝒕 𝟐 Assumption 3: 𝐯𝐞𝐜(𝑊𝛿) = 𝐹 𝐻 𝑑𝑖𝑎𝑔 𝐹𝑤 𝛿 𝐹
  • 12. MMSE channel estimation 𝑾 𝑭𝑬 𝑪 = 𝑒𝑥𝑝(𝑡𝑟 𝐹 𝐻 𝑾 𝟎 𝐹 𝐶 + 𝑇𝑙𝑜𝑔 𝐼 − 𝑾 𝑭𝑬 ) 𝐹 𝐻 𝑾 𝟎 𝐹𝑝(𝛿) 𝑑𝛿 𝑒𝑥𝑝(𝑡𝑟 𝐹 𝐻 𝑾 𝟎 𝐹 𝐶 + 𝑇𝑙𝑜𝑔 𝐼 − 𝑾 𝑭𝑬 ) 𝑝(𝛿) 𝑑𝛿 𝑝(𝛿)Distribution of channel parameters, 𝑦𝑡 𝜎2Know noise variance, Received signal, Sample covariance matrix (estimated) MMSE filter (estimated) ℎ = 𝐹 𝐻 𝑾 𝑭𝑬 𝑪 𝐹𝑦 Estimated channel Fast MMSE Estimator: Generated channel, 𝑾 𝑭𝑬 = 𝐶𝑐(𝐶𝑐 + 𝛴)−1𝐶 𝛿 = 𝑐𝑜𝑣(ℎ 𝑡) Channel covariance matrix MMSE filter coefficients ℎ 𝑡Generated channel, 𝑪 𝒄 = 𝑐𝑖𝑟𝑐(𝐶 𝛿) Circulant approximation of channel covariance matrix 𝑾 𝟎 = 𝐹𝑾 𝑭𝑬 FFT of filter 𝑪 = 𝟏 𝝈 𝟐 𝒕=𝟏 𝑻 𝐹𝒚 𝒕 𝟐 𝑊𝑆𝐸 𝑐𝑜𝑚𝑝𝑢𝑡𝑎𝑡𝑖𝑜𝑛 → 𝛰(𝑀log𝑀)
  • 13. MMSE channel estimation 𝑾 𝑭𝑬 𝑪 = 𝑒𝑥𝑝(𝑡𝑟 𝐹 𝐻 𝑾 𝟎 𝐹 𝐶 + 𝑇𝑙𝑜𝑔 𝐼 − 𝑾 𝑭𝑬 ) 𝐹 𝐻 𝑾 𝟎 𝐹𝑝(𝛿) 𝑑𝛿 𝑒𝑥𝑝(𝑡𝑟 𝐹 𝐻 𝑾 𝟎 𝐹 𝐶 + 𝑇𝑙𝑜𝑔 𝐼 − 𝑾 𝑭𝑬 ) 𝑝(𝛿) 𝑑𝛿 𝑝(𝛿)Distribution of channel parameters, 𝑦𝑡 𝜎2Know noise variance, Received signal, Sample covariance matrix (estimated) MMSE filter (estimated) ℎ = 𝐹 𝐻 𝑾 𝑭𝑬 𝑪 𝐹𝑦 Estimated channel Motivation for Low-Complexity Neural Network: Generated channel, 𝑾 𝑭𝑬 = 𝐶𝑐(𝐶𝑐 + 𝛴)−1𝐶 𝛿 = 𝑐𝑜𝑣(ℎ 𝑡) Channel covariance matrix MMSE filter coefficients ℎ 𝑡Generated channel, 𝑪 𝒄 = 𝑐𝑖𝑟𝑐(𝐶 𝛿) Circulant approximation of channel covariance matrix 𝑾 𝟎 = 𝐹𝑾 𝑭𝑬 FFT of filter 𝑪 = 𝟏 𝝈 𝟐 𝒕=𝟏 𝑻 𝐹𝒚 𝒕 𝟐 𝑭 𝑯 𝒅𝒊𝒂𝒈 𝑭𝒂 𝑭𝒙 = 𝒂 ∗ 𝒙
  • 14. MMSE channel estimation Learning Fast MMSE Estimator using a Convolutional Neural Network: 𝑦𝑡 𝜎2Know noise variance, Received signal, ℎ = 𝐹 𝐻 𝑾 𝑪𝑵𝑵 𝐹𝑦 Estimated channel ℎ 𝑡Generated channel, Conv ReLU Conv Bias Bias 𝑾 𝑪𝑵𝑵 𝑦𝑡Generated channel, Parameter Value Epochs 10 000 Training batch size 600 Testing batch size 100 Learning rate (64/M)*1e-4 Convolutional layer 2 Hierarchical Training 1. Train CNN for M antennas 2. Interpolate kernel from length M to 2M 3. Use interpolated values of kernel from previous step for initializing CNN when training for M antennas.
  • 15. MMSE channel estimation Learning Fast MMSE Estimator using a Convolutional Neural Network: ℎ 𝑡Generated channel, Conv ReLU Conv Bias Bias 𝑾 𝑪𝑵𝑵 𝑦𝑡Generated channel,
  • 16. MSE per antenna at an SNR of 0 dB for estimation from a single snapshot (T=1). Channel model with one propagation path with uniformly distributed angle.
  • 17. MSE per antenna at an SNR of 0 dB for estimation from a single snapshot (T=1). Channel model with three propagation paths.
  • 18. MSE per antenna for M = 64 antennas and for estimation from a single snapshot (T = 1). Channel model with three propagation paths.