Browsing by Author "Ojenge, Winston"

Now showing 1 - 2 of 2
  • Thumbnail Image
    Item
    PSO of Neural Networks to Predict Busy Times of Cellular Traffic for Assignment to TV Idle Channels by Cognitive Radio
    (IEEE, 2013) Ojenge, Winston; Afullo, Thomas; Ogao, Patrick; Okello Odongo, William
    Kenya has identified radio spectrum as a key driver in its development. Yet, globally, radio spectrum is inefficiently utilized due to ITU’s static spectrum allocation. In Kenya, mobile operators are running short of bandwidth due to deployment of 4G services, which enable superfast mobile broadband/internet. In the USA and UK, FCC and Ofcom, respectively, have made effort to allow opportunistic ‘poaching’ of licensed spectrum as long as communication of licensed user is not interfered with. This has focused research on use of cognitive radio, which would use its sensor networks to establish which TV channels are idle in order to allocate them temporarily to cellular networks. Enabling the cognitive radio to predict which channels shall lie idle at what times introduces better planning and more temporally-efficient allocation. This study explores the viability of predicting the times of mobile telephony traffic jam for a mobile service operator with poor QoS rating within a cell of perennial mobile traffic jam in order to explore whether those times can map well with the TV spectrum holes. The times of the TV spectrum holes shall be determined in a later study.
  • Thumbnail Image
    Item
    Use of GA-Optimized NN to Predict DVB-T2 Receiver Spectrum Holes to Accommodate Burden GSM Voice Calls
    (International Journal of Simulation--Systems, Science & Technology, 2015) Ojenge, Winston; Okelo-Odongo, William; Ogao, Patrick
    Cognitive radio enables opportunistic utilization of under-used spectrum for networks that are overwhelmed. In Nairobi city, mobile telephony networks are overwhelmed while broadcast TV channels lie idle in some parts. Research in cognitive radio concentrates on point-to-point communication and successfully conducts transmitter sensing in order to establish spectrum holes. However, in broadcast communication such as terrestrial TV, transmitter detection is inefficient, as transmitter signals may be present yet that licensed channel is not tuned-into by TV receivers, rendering those frequencies essentially idle. Our paper describes an attempt to use a novel remote technique to detect, model and predict which DVB-T2 channels are not tuned-into by any TV receiver during the worst mobile traffic jam time, within an overwhelmed mobile cell. An MLP that was GA-optimized to an MSE of 0.0046245 indicated predictability of TV-viewing and predicted Channel 514 MHz as being idle during the 5.00-5.03 pm jam time slot.