In Hierarchical Modulation (HM), the binary data is partitioned into a “ high-priority” (HP) and a “ lowpriority” (LP) bit stream that are separately and independently encoded before being mapped on non-uniformly spaced constellation points. Indeed, HM is a multi-level modulation scheme, proposed for enabling unequal error protection. The HP and LP streams vary in their susceptibility to noise. The basic HP signal is more robust, in other words, heavily protected against noise and interference, whereas the LP stream has much less robustness. In this paper, we propose an improved iterative receiver based on variational Bayesian expectation-maximization (VB-EM) scheme for semi-blind joint channel estimation and data detection based on bit-interleaved coded modulation (BICM), with the aim of improving the bit error rate (BER) of the noisy LP stream. Moreover, the proposed receiver allows us to reduce the pilot symbol overhead compared to the classical HM receivers based on pilot-assisted channel estimation.