MODELS OF ANALYSIS AND RECOGNITION SYSTEM OF MUSICAL COMPOSITIONS

Abstract

The article discusses the models of the recognition system for musical compositions in the system of synthesis and analysis of musical sounds, aimed at increasing the identification capabilities of an automated system. The recognition module for musical compositions is oriented to the server part of the system, which, regardless of the client, contains a database with fingerprints of musical compositions. With the help of melody recognition algorithms with a given argument in the form of a fingerprint, the server returns a list of musical compositions that most satisfy the search conditions. The client part interacts with the server part using the developed application programming interface, which, in addition to the well-known functional approach to the REST network protocol architecture based on the HTTP protocol, where the client uses requests only in the format defined by the server part specification, also provides for the implementation of the capabilities of the network architecture approach interactions using the GraphQL query language, which allows to build query parameters on the client side. The local database contains a set of fingerprints and metadata about musical compositions to speed up the recognition process, since it allows to cover most application scenarios with the most popular musical compositions with the ability to quickly return results after local data matching without the need for client-server interaction delays. The database synchronization module is responsible for the timely updating of the local database with new fingerprints from the server side and for the general synchronization of the client and server parts of the system. The synchronization scheduler provides a synchronization schedule for the local and server databases, as well as implements strategies for optimizing battery usage and ensures work with a low level of Internet connection. The features of storing the fingerprint database of compositions and strategies for working with devices based on mobile platforms, in particular, for the Android operating system in Doze mode, which prohibits the background operation of the device at rest, except for short periods of time, the so-called support windows, which are determined by the operating system work process and allow to perform short-term background operations. Performed the optimization of the process of using the battery energy of a mobile device while synchronizing the metadata of musical compositions between the client and server components of the system.