The use of centrifugal compressors has been increasing tremendously in the last decade as they are a key component in the present energy scenario both in the modern internal combustion engine design and in advanced cycles and innovative plant layouts as fuel cell systems. Instability phenomena limit the operating range of the whole compressor system, especially during fast transients. The target is therefore to extend the minimum flow limit in order to improve the operability of each unit while avoiding compressor surge operation and guaranteeing safe operation. For this reason, it is necessary to develop a monitoring system capable of preventing surge and extending the operating range of these machines, their performance, and reliability to allow the integration with the other plant components. The experimental investigation carried out at the University of Genoa turbocharger test facility and presented in this work, consists of steady-state and transient measurements used to characterize and identify compressor behavior in correspondence of surge inception conditions to determine different techniques which could represent surge precursors. The data analysis concentrates on pressure and vibro-acoustic signals by applying different signal processing techniques in the time and frequency domain to classify compressor operation as stable or unstable. The cross correlation function and wavelet analysis have been identified as techniques to define a precursor able to detect incipient surge conditions. Through cross correlation function analysis, it has been possible to identify the presence of propagation phenomena in the system and to evaluate how these events become more significant near an unstable low-mass flow rate condition. At low mass flow rate condition, spikes of significant amplitude are well detectable in the cross correlation function indicating the rise of significant random content in the system responses associated with the rise of incipient surge condition. Additionally, the continuous wavelet transform has been applied to operational signals to show how their time-dependent spectral structure responses can highlight the rise of unstable phenomena, not easily identifiable with traditional signal processing techniques. Exploiting its features in terms of good frequency and time resolution allowed us to identify different contents in system responses regarding phenomena that take place close to surge line and were able to detect their nature in conditions very close to deep surge ones (e.g., rotating stall with its intermitting characteristic nature). Moreover, system response was studied in high frequency range and through a demodulation technique, it was found how blade passes frequency energy content change interacting with rotating stall inception, moving close to surge line. The obtained results provide an interesting diagnostic and predictive solution to detect compressor instabilities at low mass flow rate operating conditions and to prevent compressor fails.