INTRODUCTION: Blood pressure (BP) serves as a crucial parameter in the management of three prevalent chronic diseases, hypertension, cardiovascular diseases, and cerebrovascular diseases. However, the conventional sphygmomanometer, utilizing a cuff, is unsuitable for the approach of mobile health (mHealth). METHODS: Cuffless blood pressure measurement, which eliminates the need for a cuff, is considered a promising avenue. This method is based on the relationship between pulse arrival time (PAT) parameters and BP. In this study, pulse transit time (PTT) was derived from ballistocardiograms (BCG) and impedance plethysmograms (IPG) obtained from a weight-fat scale. This study aims to address two challenges using deep learning and machine learning technologies: first, identifying BCG and IPG signals with good quality, and then extracting PTT parameters from them to estimate BP. A stacked model comprising a one-dimensional convolutional neural network (1D CNN) and gated recurrent unit (GRU) was proposed to classify the quality of BCG and IPG signals. Seven parameters, including calibration-based and calibration-free PTT parameters and heart rate (HR), were examined to estimate BP using random forest (RF) and XGBoost models. Seventeen healthy subjects participated in the study, with their BP elevated through exercise. A digital sphygmomanometer was employed to measure BP as reference values. Our methodology was validated using data collected from our custom-made device. RESULTS: The results demonstrated a signal quality classification accuracy of 0.989. Furthermore, in the five-fold cross-validation, Pearson correlation coefficients of 0.953 ± 0.007 and 0.935 ± 0.007 were achieved for systolic BP (SBP) and diastolic BP (DBP) estimations, respectively. The mean absolute differences (MADs) of XGBoost model were calculated as 3.54 ± 0.34 and 2.57 ± 0.17 mmHg for SBP and DBP, respectively. DISCUSSION: The proposed method significantly improved the accuracy of cuffless BP measurement, indicating its potential integration into weight-fat scales as an unconstrained device for effective utilization in mHealth applications.