Timbral Analysis and Recording Parameter Transformations of Snare Drums

A snare drum is capable of producing a wide range of timbres influenced by playing technique, its physical construction, and the recording methods used. When a recording engineer configures drums and studio equipment, they adjust a plethora of real-world recording parameters to achieve the desired timbre. These recording parameters impart their own timbral properties by varying amounts, and in most cases the only way to modify these properties is to re-record the audio with changes applied to the real-world variables. This thesis examines methods for computational transformations of snare drum recordings to elicit perceptual changes that mimic modification of real-world recording variables. This is achieved through four main investigations, presented throughout this thesis, two which cover timbral analysis of snare drum recordings, and two which explore post-hoc recording parameter transformations. Strike velocity and microphone selection are factors known to affect snare drum timbre, the first study analyses timbral differences associated with snare drum strike velocity. Results show that listeners are able to distinguish between high and low velocity strikes using timbral cues alone, with microphone selection having no influence on this perceptual identification. Audio analysis reveals distinct temporal and spectral features, with higher velocity strikes producing greater energy in the lower mid-range and significantly longer decay times. The second study aims to demystify the subjective preference of different microphones for snare drum recording. For the majority of microphones, preference does not change between isolated strikes and those with the presence of bleed from the hi-hat and kick drum. On average, preference is higher for condenser microphones compared to dynamic. Additionally, spectral centroid and an objective measure of brightness positively correlate with subjective scores. The ability to perceptually modify drum recording parameters in a post-recording process would be of great benefit to engineers limited by time or equipment. The first post-hoc recording parameter transformation study focuses on microphone selection, mapping the spectral features from highly-preferred microphones onto a microphone with less favourable timbral characteristics. This investigation also details the development and evaluation of a robotic drum arm for consistent strike velocity. Subjective assessment reveals that participants show no preferences between recordings from highly-preferred microphones and those from a transformed least-preferred microphone. The last study employs a data-driven approach for post-recording modification of dampening and microphone position. The system consists of a autoencoder that analyses an audio input and predicts optimal parameters of one or more third-party audio effects, which process the audio to produce the desired transformations. Two novel audio effects are proposed and compared against existing audio plugins. Perceptual quality of transformations is assessed through a subjective listening test and an object evaluation is used to measure system performance, positive results demonstrate a capacity to emulate snare dampening