Ambient backscatter communications are a critical enabling technology of radio frequency identification in Internet of Things due to its full usage of ambient signals with low cost to transmit tag signals. However, the signal detection of tag signals is a challenging task as tag signals are usually superposed on the ambient radio frequency signals and the channel state information is unavailable due to its huge and unacceptable computation complexity at tags. To solve this, we in this article proposed an enhanced signal detection scheme for the differential energy detection in ambient backscatter communications. We discovered that the single detection threshold in classic detection will not provide desirable results in smart storage scenarios. We proposed a double-threshold detection method which subtly integrates differential energy detection with maximum likelihood method to improve the detection accuracy while reducing the complexity of the reader. In addition, we derived the closed-form expressions for bit error rate (BER), miss detection, and false alarm probabilities of the data bits sent by the tag. Simulation results are presented to show that our proposed detector can offer lower BER and computational complexity compared with single threshold detection schemes in low-power and low-storage scenarios, especially when the system is in a poor channel status environment.