Algorithms for channel estimation and spectrum sensing with implementation on software defined radio

Abstract

Rapid increase in wireless applications and services pose a challenge on the limited licensed spectrum. On the other hand, researchers have found that majority of the licensed spectrum bands in di erent parts of the world are under-utilized in time, frequency and space. Several researchers around the world through measurements have shown that most of the terrestrial broadcast television (TV) bands in very high frequency (VHF) and ultra-high frequency (UHF) are unoccupied. These vacant/unused channels are known as TV white space (TVWS). The powerful propagation characteristics of VHF/UHF band signals makes these bands suitable for use in rural areas where wired infrastructure is not economical to install, and the lineof- sight wireless solutions are unreliable due to vegetation, nature and man-made features. In 2004, the IEEE 802.22 working group was created to propose protocols for VHF/UHF bands. In September 2006, the working group published functional requirements document for the wireless regional area network (WRAN) system for utilization of TVWS where the WRAN devices act as secondary users (SUs) and digital TV (DTV) acts as primary users (PUs). Hence in this thesis work, channel estimation and spectrum sensing for IEEE 802.22 WRAN (although not limited) standard has been done while considering the challenges posed by TVWS systems and validation of theoretical research on real-time standard implementations. IEEE 802.11ah is also an emerging standard based on orthogonal frequency division multiplexing (OFDM). This standard is introduced for IoT application at sub-1 GHz license-exempt bands. Since large number of IoT devices cause high interference, a receiver structure is proposed for IEEE 802.11ah (although not limited) standard in the presence of interference. The uniqueness of the work is in proposing new algorithms, comparing with existing algorithms and developing analytical insights. The proposed algorithms' performance on practical systems is tested by building a test setup and doing over-the-air real-time testing. Since the IEEE 802.22 channel is sparse in nature; hence sparse channel estimation algorithms have been proposed in the presence of both Gaussian and non- Gaussian noise. The presence of co-channel and adjacent channel interference in addition to the additive white Gaussian noise can be modeled as non-Gaussian noise. An iterative time-domain based algorithm is proposed for sparse channel estimation in the presence of Gaussian noise. Natural gradient non-parametric maximum likelihood (NG-NPML) algorithm is then proposed for sparse channel estimation in the presence of non-Gaussian noise (due to the presence of co-channel and adjacent channel interference, and impulsive noise). The NG-NPML algorithm converges much faster than the classical stochastic gradient (SG) based NPML. However, the mean square error (MSE) oor is same for both the SG-NPML and the proposed NGNPML. Thus, to further improve the MSE oor of NG-NPML, an l1 norm penalty is introduced in the NG-NPML cost function. This l1 norm penalty introduces a zero-attractor (ZA) term in the NG-NPML weight update recursion which shrinks the coe cients of inactive taps and hence reduces the steady state MSE oor. In addition, the rst and second order convergence analysis of both the NG-NPML and ZA-NG-NPML are also derived. The fast convergence of NG-NPML over SG-NPML has been validated by doing the world's rst implementation of IEEE 802.22 PHY on National Instruments Universal Software Radio Peripheral (NI-USRP) 2952R in the presence of another IEEE 802.22 transmitter and DTV transmitter as co-channel interference.Rapid increase in wireless applications and services pose a challenge on the limited licensed spectrum. On the other hand, researchers have found that majority of the licensed spectrum bands in di erent parts of the world are under-utilized in time, frequency and space. Several researchers around the world through measurements have shown that most of the terrestrial broadcast television (TV) bands in very high frequency (VHF) and ultra-high frequency (UHF) are unoccupied. These vacant/unused channels are known as TV white space (TVWS). The powerful propagation characteristics of VHF/UHF band signals makes these bands suitable for use in rural areas where wired infrastructure is not economical to install, and the lineof- sight wireless solutions are unreliable due to vegetation, nature and man-made features. In 2004, the IEEE 802.22 working group was created to propose protocols for VHF/UHF bands. In September 2006, the working group published functional requirements document for the wireless regional area network (WRAN) system for utilization of TVWS where the WRAN devices act as secondary users (SUs) and digital TV (DTV) acts as primary users (PUs). Hence in this thesis work, channel estimation and spectrum sensing for IEEE 802.22 WRAN (although not limited) standard has been done while considering the challenges posed by TVWS systems and validation of theoretical research on real-time standard implementations. IEEE 802.11ah is also an emerging standard based on orthogonal frequency division multiplexing (OFDM). This standard is introduced for IoT application at sub-1 GHz license-exempt bands. Since large number of IoT devices cause high interference, a receiver structure is proposed for IEEE 802.11ah (although not limited) standard in the presence of interference. The uniqueness of the work is in proposing new algorithms, comparing with existing algorithms and developing analytical insights. The proposed algorithms' performance on practical systems is tested by building a test setup and doing over-the-air real-time testing. Since the IEEE 802.22 channel is sparse in nature; hence sparse channel estimation algorithms have been proposed in the presence of both Gaussian and non- Gaussian noise. The presence of co-channel and adjacent channel interference in addition to the additive white Gaussian noise can be modeled as non-Gaussian noise. An iterative time-domain based algorithm is proposed for sparse channel estimation in the presence of Gaussian noise. Natural gradient non-parametric maximum likelihood (NG-NPML) algorithm is then proposed for sparse channel estimation in the presence of non-Gaussian noise (due to the presence of co-channel and adjacent channel interference, and impulsive noise). The NG-NPML algorithm converges much faster than the classical stochastic gradient (SG) based NPML. However, the mean square error (MSE) oor is same for both the SG-NPML and the proposed NGNPML. Thus, to further improve the MSE oor of NG-NPML, an l1 norm penalty is introduced in the NG-NPML cost function. This l1 norm penalty introduces a zero-attractor (ZA) term in the NG-NPML weight update recursion which shrinks the coe cients of inactive taps and hence reduces the steady state MSE oor. In addition, the rst and second order convergence analysis of both the NG-NPML and ZA-NG-NPML are also derived. The fast convergence of NG-NPML over SG-NPML has been validated by doing the world's rst implementation of IEEE 802.22 PHY on National Instruments Universal Software Radio Peripheral (NI-USRP) 2952R in the presence of another IEEE 802.22 transmitter and DTV transmitter as co-channel interference.