Multi-view clustering remains a critical challenge in unsupervised learning, especially in scenarios with heterogeneous feature representations. Existing models often rely on either shallow representations or over-parameterized deep architectures that are difficult to optimize and generalize. In this paper, we propose EMV-FCRL, a novel multi-view fuzzy clustering framework that integrates representation learning and inter-view consistency through co-embedding. EMV-FCRL learns latent representations specific to each view, aligns them through embedding regularization, and jointly updates fuzzy memberships in a unified objective. An alternating optimization algorithm is designed to update the embeddings, centroids, and memberships iteratively. Extensive experiments on benchmark datasets demonstrate that EMV-FCRL consistently outperforms state-of-the-art baselines in terms of clustering accuracy, fuzzy partition quality, and inter-view alignment.