Abstarct: Signal analysis in the time-frequency domain provides more advantageous than analyzing the signal in either the time or the frequency domain. Adaptive directional time-frequency distributions (ADTFDs) are effective and superior to other fixed and adaptive methods. In the ADTFD the direction of the kernel is adapted on the basis of directional Gaussian or double derivative Gaussian filter. Although ADTFDs are very effective, their applicability are restricted due to the manual tuning of the parameters, high computational cost, sensitivity to the noise, weak performance in analyzing multicomponent signal with different amplitudes and instantaneous frequency (IF) estimation of multicomponent signals with overlapped components. In this work to overcome the above-mentioned challenges, first the parameters of the ADTFDs are estimated. Using a two-stage algorithm, first the length of the smoothing kernel is optimized globally. In the second stage, the parameters which control the shape of the selected smoothing window is optimized, locally. To alleviate computational cost a computationally efficient variant of ADTFD is introduced. In the new ADTFD, the optimized directions are estimated using the Radon transform of the ambiguity function of the signal and the searching is eliminated to the estimated directions. In the new ADTFD, the sensitivity to the noise is also covered as eliminating the wrong directions corresponding to the noise, alleviates the sensitivity of the ADTFD to the presence of the noise. By automating the ADTFD, the possibility of precise analysis of multicomponent signal with different components is provided. To estimate the IF of a multicomponent signal specifically with overlapped components, a new method baxsed on the proposed time-frequency technique and time-frequency filtering is introduced. The performance of the proposed method in contrast to the state of the art methods such as ICCD-RPRG is compared. baxsed on the proposed IF estimate method a new algorithm for reconstruction of multicomponent signals with messing samples is introduced. The performance of this method is compared with the common methods such as gradient descent, EOMP and iterative thresholding with different levels of missing samples, which leads to minimum MSE for the proposed method. As part of the case study, the performance of the proposed ADTFD is assessed using a seizure detection system and a spike detection algorithm for signals with high frequency activity and low SNR. In case of seizure detection system, the obtained results show classification accuracy of 98.56%, which is 37% more than the accuracy achieved with other TFDs. In order to detect the spike, a new algorithm baxsed on the direction of the energy of the signal component in the time-frequency plane is introduced. The performance of the proposed spike detection method is compared with SNEO and CoB methods using different statistical measures such as precision rate and hit-rate.