• Literature study for research done in the relevant domain
• In-depth analysis of a variety of algorithms for pre-processing, feature and classification and their qualitative analysis for application to the problem domain
• Definition of critical parameters for automated signal feature detection and classification system
• Identification of possible signal noise artefacts and corresponding algorithmic schemes
• Data analysis using Octave (Open source equivalent of Matlab)

Customer

The customer is a leading manufacturer of blood analysis devices. For a new generation blood analysis device, the customer wanted to get an algorithmic solution to address stringent automated detection of signal features for a wider variety of samples.

Agiliad Solution

The solution involved breaking down system into modules and identifying algorithmic options for each module. Critical success factors for signal pre-processing, feature detection and classification were identified and how each algorithm can address challenges was analysed. Finally detailed information about for each algorithm for each module was provided for actual implementation. Some of the Algorithms analysed were,

• Pre-processing : Baseline correction, Noise filter, Normalization
• Feature Detection : Gradient based peak detection, Peak detection using Continuous wavelet transform, Template matching using matched filter
• Pattern Classification : Heuristic or rule based classifier, Statistical classifier

Agiliad value proposition and customer benefits

• Option value comparison of various design options
• Concept to realization with innovative application of algorithms
• Extensible design for possible product feature additions
• Excessive references from literature and prior-art
• Techniques for data analysis using different data visualization techniques