Identification of bacterial species and strain currently relies on either lengthy culture-based methods, or complex methods involving lengthy sample preparation or the introduction of exogenous labels. As the identification of bacteria in clinical and homeland security applications is often time-sensitive, the need for methods that maintain high levels of accuracy whilst offering improvements in terms of speed and simplicity of analysis is growing rapidly. Enhanced Raman techniques are becoming increasingly common research tools in the biosciences thanks to their non-destructive, speed of analysis, and label-free nature. The work presented here demonstrates the capabilities of enhanced Raman techniques outside of research. The studies in this thesis explore the identification of bacteria in cases that are more complex than pure isolates on slides, and the performance of SERS-active monolayers in comparison to the ubiquitous nanoparticle colloids that are commonly used in bacterial SERS. Herein, a layer-by-layer SERS-active substrate based on polycarbonate filters are used to capture bioaerosols and provide information about bacterial species and phenotype with analysis times below one minute. A novel method for increasing classification accuracy by combining multiple excitations to achieve a combination of resonant and non-resonant spectral data is also presented within. This method is demonstrated to achieve greater than 99% accuracy for bacterial samples, even in complex molecular environment of an artificial sputum model that is representative of the sputum of the lung of cystic fibrosis patients. Additionally, an easily synthesised self-assembled gold monolayer is presented as a point of comparison to the use of aggregated gold nanoparticles, and found to offer greater classification accuracy. This thesis shows the impressive accuracy attainable by enhanced Raman methods, and offers new tools to help further establish Raman spectroscopy and its enhanced techniques as a powerful analytical technique for both research and detection in complex real-world contexts.