This paper explores the learning capability of hidden Markov model (HMM) in capturing the temporal correlation and predicting primary user (PU) activity pattern of real spectrum data of GSM-900 band through an USRP-LabVIEW platform for cognitive radio (CR) systems. The inability of the widely used stationary Markov model in estimating the occupancy pattern of primary channels for a long duration of time has been verified. We proposed an alternative duty cycle (DC)–based two-state discrete-time Markov chain (DTMC-DC) model. Analysis of empirical data indicates that DC required for a non-stationary DTMC-DC model can be well approximated by a trapezoidal shape and the PU spectrum usage pattern estimated using DTMC-DC is capable of learning the statistical behavior (length of idle and busy interval periods) of a real channel accurately with a reduced complexity.

中文翻译：

---

信道估计中基于占空比的DTMC模型的经验验证和性能

本文探讨了隐马尔可夫模型（HMM）在通过认知无线电（CR）系统的USRP-LabVIEW平台捕获时间相关性和预测GSM-900频段真实频谱数据的主要用户（PU）活动模式方面的学习能力。事实证明，广泛使用的固定马尔可夫模型无法长时间估算主信道的占用模式。我们提出了基于替代占空比（DC）的两状态离散时间马尔可夫链（DTMC-DC）模型。