Mechanical shocks widely exist in many important engineering applications. These mechanical shocks normally cannot cause damage to the main structures of vehicles, but they may possibly result in major functional failure of on-board products, which would subsequently result in the partial or total loss of a mission. Products should be tested accordingly in laboratory environment to ensure their reliability under mechanical shock environment during their service life. However, strictly reproduction of the mechanical shock environment for products validation would be almost impossible, especially with common mechanical testing methods. Unlike other simple impulse or stationary vibration excitations, mechanical shock severity are too complicated to be understood in time and frequency domain. This thesis studies the shock severity problem with mechanical and mathematical techniques. Specifically, signal processing and data visualisation methods in both time and response domains are proposed. According to shock generation mechanism, a characteristic shock waveform is identified, with which important temporal information can be extracted from general shock measurements. By defining shock waveform as the impulse response function of a filter, a shock synthesis method is proposed to synthesize shock signals with the same severity of a given shock measurement. Extracted by singular value decomposition, the proposed shock severity infimum along with the traditional shock response spectrum (SRS) can bound the range of shock severity. Based on the linear superposition assumption, low-pass-filter-based SRS method is proposed, which improves the current SRS method by providing frequency distribution information. Methods proposed in this thesis can compare the relative severity of various shock environments, and can be used as tools to specify shock testing specifications.