Intellectual Merit |
On
the theoretical aspect, we have proved a strong result which serves as the
foundation of this project, that is, a
sinusoid can be approximated as a linear combination of base sinusoids on constant
frequencies and the approximation error decays exponentially fast
as the number of base sinusoids increases. This result basically explains the main theoretical
question of this project, and have been used as a guide for exploring design
options in a number of practical problems. The proof is mathematically
interesting and is also the first such result on the approximation of a
sinusoid by sinusoids in the literature.
On
the practical aspect, we expect our results on CSI compression and estimation
to have profound impact on the design and real-world performance of wireless
networks such as Wi-Fi and cellular networks. Our patented CSI compression algorithm, CSIApx,
achieves higher compression ratio, lower distortion, and lower computation
complexity than the existing CSI compression algorithm. With high
compression ratio, more frequent CSI feedbacks can be obtained, and more
efficient data transmission options, such as MU-MIMO, can be enabled or enjoy
lower error ratio. Our CSI estimation
algorithm, ParEst, achieves much higher accuracy than
existing algorithms in challenging cases when the transmitter has multiple
antennas. ParEst
can be used for CSI estimation in a number of scenarios, such as 5G MIMO
uplink and Grant Free channels, and improve the performance of the network.