This paper deals with spectrum sensing for cog-nitive radio-based Internet of Things (IoT) systems and theircoexistence with Long Term Evolution (LTE) systems. Due tothe sparsity of the covariance matrix of IoT/LTE signals, wereveal that the likelihood ratio test approximates to energydetection (ED) at low signal to noise ratio. However, the noise(power) uncertainty can degrade the performance of ED severely,especially when low-cost IoT devices are employed for spec-trum sensing. To tackle this issue, we derive the relationshipamong noise power, total power, and autocorrelation coefficientof received signals, and propose an unbiased estimator of noisepower without the knowledge of the presence/absence of IoT/LTEsignals. We then design a new ED with multiple estimates of noisepower from historical and current sensing data, and analyze itstheoretical performance. Numerical results are provided to verifythe theoretical results and demonstrate the superior performanceof the proposed detector. It is shown that, by exploiting sufficienthistorical sensing data, the performance of the proposed ED canclosely approach that of the ideal ED.