Bone-conducted phonocardiogram (BCPCG) recordings acquired in practical environments are often contaminated by motion artifacts and muscle tremor, making signal quality assessment essential for reliable downstream analysis. This study proposes a multi-domain handcrafted feature framework for BCPCG quality assessment based on raw signals and their envelope representations. A 25-dimensional feature set was designed to characterize signal quality from waveform morphology, energy variation, spectral structure, envelope peak organization, and rhythm consistency. Experiments on 24,581 labeled BCPCG segments were conducted under a three-class setting using five-fold stratified crossvalidation. Among six evaluated machine-learning classifiers, LightGBM achieved the best overall performance, with an ACC of 0.914 and an AUC of 0.967. The proposed framework provides an effective and compact solution for BCPCG quality assessment. Keywords: Bone-conducted Phonocardiogram; signal quality assessment; multi-domain features;