A Wavelet-Based Methodology for Features Extraction in Postural Instability Analysis

A Comparison among Different Strategies to Detect Potential Unstable Behaviors in Postural Sway

“…Figure 6 shows features (8)–(10) estimated by the DWT for each of the computed levels. As expected, the contribution of the first two levels is negligible and is not used to feed the NF algorithm [ 30 ]. Only a limited number of features are shown in the picture for the sake of readability.…”

“…The 5-level DWT of the DAP and DML were computed over time windows of 10 s shifted by 1 s to each other. Successively, the following features were estimated (Mean Value (MV), Standard Deviation (STD), and Energy (E), which the literature recognizes as the most meaningful, in order to accomplish the task under consideration [ 28 , 29 , 30 ]. where ‘ T ( a , k)’ represents a wavelet at timescale a , identified by levels d1, d2, d3, d4, d5 , and K the number of samples in the transformed signal.…”

“…In order to reduce the number of inputs of the postural detection algorithm, the approach proposed in [ 30 ] was followed, mainly considering lower levels ( d3 , d4 , and d5 ) of the Wavelet Coefficients (WCs). As an example, Section 4.2 shows that values of the strategic feature (8) assume greater amplitude for levels ( d3, d4 , and d5 ) than for levels ( d1 and d2 ).…”

“…Moreover, the same work offers a very well-elaborated background on wavelet transform analysis tailored to specific cases of use. In [ 30 ], DWT is used to process signals provided by a micro-controlled-based inertial wearable system with the aim of extracting useful information related to postural sway behavior.…”

“…Two major classes are represented by CRISP and Machine Learning (ML) approaches. In [ 30 ], the authors proposed a methodology for the postural-instability-detection-exploiting features provided by the DWT theory and a threshold-based classification strategy. In [ 22 ], features extracted by the time evolution of Antero-Posterior and Medio-Lateral Displacements were used with a threshold-based paradigm to detect potential instability in the postural behavior.…”