This paper proposes a superimposed symbol scheme for double-RIS aided integrated sensing and communication (ISAC) systems to enhance spectral efficiency (SE) while achieving accurate localization. The scheme superimposes sensing pilots onto data symbols, reducing the overhead for sensing and improving SE. A structure-aware sparse Bayesian learning (SBL) framework is developed to extract channel angles from coupled channels, using decoded data as side information to improve sensing performance. A low-complexity simultaneous communication and localization algorithm is proposed, employing unitary approximate message passing (UAMP) for initial angle estimation and iterative refinement through reduced-dimension matrix calculations. The sparse code multiple access (SCMA) technology is integrated to suppress multi-UE interference and enhance data detection. Numerical results show that the proposed scheme achieves centimeter-level localization while attaining up to 96% of the SE of conventional communications without sensing capabilities. Compared to other ISAC schemes, the proposed superimposed symbol scheme provides an effective throughput improvement of over 133%. The system model involves a multi-UE uplink ISAC scenario with two RISs, where sensing pilots are superimposed with data symbols to reduce training and localization costs. The channel model uses the Saleh-Valenzuela model for RIS-BS channels and a basis expansion model for channel angle estimation. The signal model includes data and sensing pilots transmitted by the fixed site and multiple UEs. The proposed algorithm uses Bayesian learning to estimate data symbols and channel angles, with iterative refinement through UAMP and SCMA. The results demonstrate the effectiveness of the proposed scheme in achieving high SE and accurate localization in double-RIS aided ISAC systems.This paper proposes a superimposed symbol scheme for double-RIS aided integrated sensing and communication (ISAC) systems to enhance spectral efficiency (SE) while achieving accurate localization. The scheme superimposes sensing pilots onto data symbols, reducing the overhead for sensing and improving SE. A structure-aware sparse Bayesian learning (SBL) framework is developed to extract channel angles from coupled channels, using decoded data as side information to improve sensing performance. A low-complexity simultaneous communication and localization algorithm is proposed, employing unitary approximate message passing (UAMP) for initial angle estimation and iterative refinement through reduced-dimension matrix calculations. The sparse code multiple access (SCMA) technology is integrated to suppress multi-UE interference and enhance data detection. Numerical results show that the proposed scheme achieves centimeter-level localization while attaining up to 96% of the SE of conventional communications without sensing capabilities. Compared to other ISAC schemes, the proposed superimposed symbol scheme provides an effective throughput improvement of over 133%. The system model involves a multi-UE uplink ISAC scenario with two RISs, where sensing pilots are superimposed with data symbols to reduce training and localization costs. The channel model uses the Saleh-Valenzuela model for RIS-BS channels and a basis expansion model for channel angle estimation. The signal model includes data and sensing pilots transmitted by the fixed site and multiple UEs. The proposed algorithm uses Bayesian learning to estimate data symbols and channel angles, with iterative refinement through UAMP and SCMA. The results demonstrate the effectiveness of the proposed scheme in achieving high SE and accurate localization in double-RIS aided ISAC systems.