In this paper, we address the challenging problem of high pilot overhead and peak-to-average power ratio (PAPR) in channel estimation in high-mobility scenarios for 5G and beyond. To improve the channel estimation accuracy and enhance the reliability of data detection, we propose a two-stage algorithm for channel estimation and symbol detection in OTFS systems. In the first stage, a coarse estimate of the channel is obtained, followed by interference cancellation. Then, data symbols are detected using a message-passing (MP) algorithm. In the second stage, detected data and pilot symbols are jointly used to accurately estimate the channel. Simulations show that our proposed scheme, namely contiguous embedded pilots (CEP), achieves a good trade-off between complexity and performance in terms of normalized mean square error (NMSE), bit error rate (BER), spectral efficiency, and PAPR when compared to embedded pilot (EP) and superimposed pilot (SP) designs.