In medical applications, biomedical acquisition systems (BASs) are frequently used in order to diagnose and monitor critical conditions such as stroke, epilepsy, Alzheimer disease, arrhythmias and etc. Biomedical signals (BSs), which produce valuable information about the condition of various physiological subsystems in our body, can be obtained by using multi-channel BASs. Due to the time-varying behavior of physiological sub-systems, most of the BSs are expected to have non-stationary character. In order to derive desired clinical information from these non-stationary BSs, an appropriate analysis method which exhibits adjustable time-frequency resolution is needed. The wavelet transform (WT), in which the time-frequency resolution can be adjusted according to the different parts of the signal, are widely used in the analysis of BSs. The discrete wavelet transform (DWT) is a fast and discretized implementation of classical WT and was employed as a feature extractor and de-noising operator for BSs in literature. However, due to the aliasing, lack of directionality and being shift-variance disadvantages, the DWT exhibits limited performance. A modified version of the DWT, which is named as Dual Tree Complex Wavelet Transform (DTCWT), is employed in the analysis of BSs and improved results are obtained. Therefore, in this study, considering the improvements in embedded system technology and the needs for wavelet based multi-channel real-time feature-extraction/de-noising operations in portable medical devices, the DTCWT is implemented as a multi-channel system-on-chip by using field programmable gate arrays. In proposed hardware architecture, for N input-channels, the DTCWT is implemented by using only one adder and one multiplier. The area efficiency and speed limits of proposed system are presented comparing with our previous approaches.