The simulation analysis of the proposed compressive spectrum sensing model is performed in MATLAB tool. To replicate the realistic wireless communication network the simulation model utilizes 612 subcarriers in an OFDM system. To model the multipath propagation effects, TDL-A delay profile is employed. The channels are characterized by a delay spread of 3e-7 seconds. By simulating non line of sight conditions Rayleigh fading is used so that the condition typically replicates the communication network in urban areas. In order to reflect the relative motion between transmitter and receiver, a maximum doppler shift of 50Hz is incorporated so that dynamic nature of mobile communication can be represented. The bandwidth of 14 OFDM symbols is selected as 20 MHz with subcarrier spacing of 30kHz. This setup provides a comprehensive testing environment for compressive spectrum sensing technique and closely related to real time communication environments. The parameters used in the proposed model simulation analysis are listed in Table 1 and Table 2 provides the details about the simulation environment.

Table 1

S.No | Algorithm | Parameter | Value/Range |

1 | QI-RNN | Number of Layers | 2–4 layers |

2 | Hidden Units per Layer | 64–256 units |

3 | Activation Function | ReLU, Tanh |

4 | Learning Rate | 0.001–0.01 |

5 | Batch Size | 32, 64, 128 |

6 | Epochs | 50–200 |

7 | EHO-BA | Population Size | 30–50 |

8 | Max Iterations | 100–200 |

9 | Inertia Weight (w) | 0.5–0.9 |

10 | Cognitive Coefficient (c1) | 1.5-2.0 |

11 | Social Coefficient (c2) | 1.5-2.0 |

12 | Attraction Coefficient (β) | 0.1–0.3 |

13 | Mutation Rate | 0.01–0.1 |

14 | Loudness (A) | 0.5-1.0 |

15 | Pulse Rate (r) | 0.5-1.0 |

Table 2

Simulation Environment details

S.No | Parameters | Values |

1 | Number of subcarriers | 612 |

2 | Delay Profile | TDL-A |

3 | Delay Spread | 3e-7 |

4 | Maximum Doppler Shift | 50 |

5 | Number of OFDM symbols | 14 |

6 | Channel Type | Rayleigh fading |

7 | Subcarrier Spacing | 30 |

8 | Bandwidth | 20 MHZ |

The proposed model performance is evaluated using different metrics like throughput, energy efficiency, energy consumption, false alarm probability, mean squared error (MSE), sensing accuracy, and detection probability, at different Signal-to-Noise Ratios (SNRs). To validate the better performance existing methods like Adaptive Wideband Spectrum Sensing (AWBSS) Model, Multi-task Compressive Sensing - Fast Fourier Transform Accumulation Method (MSC-FAM), Convolutional Neural Network (CNN) Based Sensing Model, and Deep belief network with sunflower optimization (DBN-SFO) are considered for comparative analysis [25].

Figure 4 depicts the throughput analysis of proposed model for SNR of 0 to 30 dB. The results depicts that the proposed QIRNN-EHOBA outperforms the existing models. The proposed model attains a maximum throughput of 6.8bits/sec/Hz at 30 dB which is significantly higher compared to existing methods. The throughput attained by DBN-SFO is 6 bits/sec/Hz, CNN is 5.5 bits/sec/Hz, MCS-FAM is 4 bits/sec/Hz and AWBSS is 3 bits/sec/Hz is comparatively lesser than the proposed QIRNN-EHOBA model throughput. The advanced quantum principles combined with the neural network and the optimized hyper parameters provided this enhanced throughput over existing methods. The improved throughput of proposed model also indicates the efficient utilization of spectral resources and improved data transmission rates in noisy conditions.

The energy efficiency analysis given in Fig. 5 depicts that the proposed model energy efficiency increases when SNR increases. The proposed model exhibited a maximum energy efficiency of 0.95 bits/Joule for SNR of 30dB. While the existing methods like DBN-SFO, CNN, and MCS-FAM saturates around 0.9 bits/Joule. The least performance is exhibited byAWBSS with energy efficiency of 0.85bits/Joule. The higher energy efficiency of proposed QIRNN-EHOBA model indicates that it effectively balances the trade-off between throughout and energy consumption. Due to this, maximal performance is attained by the proposed QIRNN-EHOBA model compared to existing methods.

The energy consumption analysis given in Fig. 6 depicts how the energy consumption changes as there is a change in SNR. From figure it can be observed that proposed model exhibits minimum energy consumption of about 12 Joules at 30dB. Whereas existing DBN-SFO consumes around 14 Joules, MCS-FAM about 15 Joules, CNN approximately consumes 19 Joules, and AWBSS consumes 13 Joules at 15 dB. This remarkable reduction in energy consumption by the propsoed QIARNN-EHOBA is due to the efficient optimization technique that reduces the energy use without compromising the performance.

The false alarm probability analysis given in Fig. 7 highlights the proposed model lowest false alarm probability compared to existing methods. The false alarm probability decreases as SNR increases and for the proposed model the false alarm probability reduced to 0.1 at 30dB. Whereas existing DBN-SFO, CNN, MCS-FAM, and AWBSS show higher false alarm probabilities of around 0.15, 0.2, 0.25, and 0.3, respectively. The proposed QIRNN-EHOBA effectively differentiates the noise and actual signals due to its efficient feature extraction and decision-making features. Thus, the false alarm probability of proposed QIRNN-EHOBA reduces greatly compared to existing methods and enhances the reliability in spectrum sensing process.

The mean square error analysis is comparatively analysed in Fig. 8 with respect to SNR. The error reduces when SNR increases. The proposed QIRNN-EHOBA model exhibits the lowest MSE, approximately 0.1 at 30 dB, whereas DBN-SFO exhbits MSE of 0.15, CNN exhibits MSE of 0.2, MCS-FAM exhibits MSE of 0.25, and AWBSS exhibits MSE of 0.35. The proposed model lower MSE highlights the proposed model superior accuracy in signal estimation. The advanced learning procedure of proposed model effectively differentiates noise and ensures precise and reliable performance.

The sensing accuracy of the proposed model is comparatively analyzed with existing methods in Fig. 9. The proposed model achieves highest sensing accuracy of 0.95 at 30dB. Whereas existing methods attains lowest accuracies of around .9, 0.88, 0.85, and 0.8, respectively forDBN-SFO, CNN, MCS-FAM, and AWBSS methods. The highest sensing accuracy of proposed model highlights its ability in classifying signals under various SNR conditions which is highly effective for cognitive 5G networks.

The detection probability analysis of proposed model is comparatively analysed and presented in Fig. 10 with respect to SNR. The proposed QIRNN-EHOBA exhibits its higher detection probability of 0.9 at 30dB which is superior than the detection probability of existing DBN-SFO of 0.85, CNN of 0.83, MCS-FAM of 0.8 and AWBSS of 0.75. The efficient hyperparameter optimization of the proposed model ensures reliable detection and enhances the overall detection probability compared to existing methods.

The comparative analysis given in Fig. 11 (a) depicts the performance of proposed model throughput, energy efficiency and false alarm probability for different SNR values of 0 dB, 10 dB, and 20 dB. The proposed model exhibits highest throughput for all the SNR values compared to existing methods. The proposed QIRNN-EHOBA exhibits 3.5 bits/sec/Hz at 0 dB, 5 bits/sec/Hz at 10 dB, and 8 bits/sec/Hz at 20 dB. Whereas existing DBN-SFO exhibits about 3 bits/sec/Hz at 0 dB, 4.5 bits/sec/Hz at 10 dB, and 6.5 bits/sec/Hz at 20 dB, CNN exhibit 2.5 bits/sec/Hz at 0 dB, 4 bits/sec/Hz at 10 dB, and 6 bits/sec/Hz at 20 dB. The existing MCS-FAM and AWBSS models perform lower than the proposed model specifically AWBSS performs lower among all in terms of throughput for different SNR values.

The energy efficiency proposed model and existing models with SNR values of 0 dB, 10 dB, and 20 dB depicted in Fig. 11 (b) exhibits the higher energy efficiency of proposed model over existing methods. The proposed QIRNN-EHOBA model steadily attains the highest energy efficiency across all SNR values specifically about 1 bits/Joule at 20 dB. Whereas existing DBN-SFO and CNN exhibits energy efficiencies nearly 0.9 bits/Joule at 20 dB. In terms of energy efficiency also, the performance of AWBSS and MCS-FAM is lower than the proposed QIRNN-EHOBA model.

The false alarm probability of proposed model and existing models with SNR values of 0 dB, 10 dB, and 20 dB depicted in Fig. 11 (c) highlights the lowest false alarm probability of proposed model. The proposed model exhibits 0.2 at 0 dB and it reduced to less than 0.05 at 20 dB. Whereas existing DBN-SFO and CNN exhibits false alarm probabilities of around 0.1 at 20 dB. The existing MCS-FAM and AWBSS exhibits higher false alarm probabilities compared to proposed model.

From the experimental analysis the proposed model for spectrum sensing in cognitive 5G networks is demonstrated. The experimentations confirmed the proposed model superior accuracy over the existing methods in terms of throughput, energy efficiency, energy consumption, false alarm probability, mean squared error (MSE), sensing accuracy, and detection probability, at different Signal-to-Noise Ratios (SNRs). Specifically proposed model attained throughput of 6.8 bits/sec/Hz at 30 dB, energy efficiency of 0.95 bits/Joule at 30dB, lowest energy consumption of 12 Joules at 30 dB, false alarm probability of 0.1 at 30 dB, sensing accuracy and detection probability of 0.95 and 0.9at 30 dB. Overall, the proposed QIRNN-EHOBA with advanced learning algorithm and optimization technique improves the effectiveness of spectrum sensing in cognitive 5G networks over existing methods.