Implement the line spectral estimation from quantized measurements. This code (VALSE-EP) is built based on the VALSE algorithm, and is written by Jiang Zhu, Qi Zhang and Xiangming Meng. The VALSE-EP is designed to perform line spectral estimation from quantized samples. If you have any problems about the code, please feel free to contact jiangzhu16@zju.edu.cn
Please directly run NMSEvsIter.m to see the performance of the VALSE-EP. In addition, the VALSE code (written by Mihai-Alin Badiu) is also provided for performance comparison.
Main function： out = VALSE_EP( y_q, m, ha, x, Iter_max, B, yy_min, alpha, method_EP )
Input parameters：
y_q：For quantization, it belongs to
m: The index correspond to incomplete measurements
ha: Set ha=2, which corresponds to Heuristic 2
x: True spectral signal
Iter_max：The maximum number of iterations
B: The bit-depth. For unquantized setting, set B=inf.
yy_min: The left endpoints of the quantizer
alpha: The stepsize of the quantizer
method_EP：'diag_EP' or scalar_EP’. Please set 'diag_EP'.
Output parameters:
out=struct('freqs',th,'amps',w(s),'x_estimate',xr,'noise_var',nu,'iterations',t,'MSE',mse,'K',Kt);
freqs: The point estimates of the frequency
amps: The complex weight amplitude
xr：Reconstructed line spectral
noise_var: noise variance estimate
t: The iterations exited
MSE：Normalized MSE
K: Tracking the number of spectral during iteration