This dissertation aims at improving spectrum sensing algorithms in order to effectively apply them to cellular systems. In wireless communications, cellular systems occupy a significant part of the spectrum. The spectrum usage for cellular systems are rapidly expanding due to the increasing demand for wireless services in our society. This results in radio frequency spectrum scarcity. Cellular systems can effectively handle this issue through cognitive mechanisms for spectrum utilization. Spectrum sensing plays the first stage of cognitive cycles for the adaptation to radio environments. This dissertation focuses on maximizing the reliability of spectrum sensing to satisfy regulation requirements with respect to high spectrum sensing performance and an acceptable error rate. To overcome these challenges, characteristics of noise and manmade signals are exploited for spectrum sensing. Moreover, this dissertation considers system constraints, the compatibility with the current and the trends of future generations. Newly proposed and existing algorithms were evaluated in simulations in the context of cellular systems. Based on a prototype of cognitive cellular systems (CCSs), the proposed algorithms were assessed in realistic scenarios. These algorithms can be applied to CCSs for the awareness of desired signals in licensed and unlicensed bands. For orthogonal frequency-division multiplexing (OFDM) signals, this dissertation exploits the characteristics of pilot patterns and preambles for new algorithms. The new algorithms outperform the existing ones, which also utilize pilot patterns. Additionally, the new algorithms can work with short observation durations, which is not possible with the existing algorithms. The Digital Video Broadcasting - Terrestrial (DVB-T) standard is taken as an example application for the algorithms. The algorithms can also be developed for filter bank multicarrier (FBMC) signals, which are a potential candidate for multiplexing techniques in the next cellular generations. The experimental results give insights for the reliability of the algorithms, taking system constraints v into account. Another new sensing algorithm, based on a preamble, is proposed for the DVBT2 standard, which is the second generation of of DVB system. DVB-T2 systems have been deployed in worldwide regions. This algorithm can detect DVB-T2 signals in a very short observation interval, which is helpful for the in-band sensing mode, to protect primary users (in nearly real-time) from the secondary transmission. An enhanced spectrum sensing algorithm based on cyclostationary signatures is proposed to detect desired signals in very low signal-to-noise ratios (SNRs). This algorithm can be developed to detect the single-carrier frequency division multiple access (SC-FDMA) signal, which is adopted for the uplink of long-term evolution (LTE) systems. This detector substantially outperforms the existing detection algorithms with the marginal complexity of some scalar multiplications. The test statistics are explicitly formulated in mathematical formulas, which were not presented in the previous work. The formulas and simulation results provide a useful strategy for cyclostationarity-based detection with different modulation types. For multiband spectrum sensing, an effective scheme is proposed not only to detect but also to classify LTE signals in multiple channels in a wide frequency range. To the best of our knowledge, no scheme had previously been described to perform the sensing tasks. The scheme is reliable and flexible for implementation, and there is almost no performance degradation caused by the scheme compared to single-channel spectrum sensing. The multiband sensing scheme was experimentally assessed in scenarios where the existing infrastructures are interrupted to provide mobile communications. The proposed algorithms and scheme facilitate cognitive capabilities to be applied to real cellular communications. This enables the significantly improved spectrum utilization of CCSs.