Biblio

Most of existing audio fingerprinting systems have limitations to be used for high-specific audio retrieval at scale. In this work, we generate a low-dimensional representation from a short unit segment of audio, and couple this fingerprint with a fast maximum inner-product search. To this end, we present a contrastive learning framework that derives from the segment-level search objective. Each update in training uses a batch consisting of a set of pseudo labels, randomly selected original samples, and their augmented replicas. These replicas can simulate the degrading effects on original audio signals by applying small time offsets and various types of distortions, such as background noise and room/microphone impulse responses. In the segment-level search task, where the conventional audio fingerprinting systems used to fail, our system using 10x smaller storage has shown promising results. Our code and dataset are available at https://mimbres.github.io/neural-audio-fp/.

In this article analysis of the effectiveness of the acoustic algorithm of the fingerprint in the conditions of various acoustic disturbances is presented and described. The described algorithm is stable and should identify music even in the presence of acoustic disturbances. This was checked in a series of tests in four different conditions: silence, street noise, noise from the railway station, noise from inside the moving car during rain. In the case of silence, 10 measurements were taken lasting 7 seconds each. For each of the remaining conditions, 21 attempts were made to identify the work. The capture time for each of the 21 trials was 7 seconds. Every 7 attempts were changed noise volume. Subsequently, they were disruptions at a volume lower than the volume of the intercepted song, another 7 with an altitude similar to the intercepted track, and the last with a much higher volume. The effectiveness of the algorithm was calculated for two different times, and general - for the average of two results. Base of "fingerprints" consisted of 20 previously analyzed music pieces belonging to different musical genres.

Acoustic microscopy is characterized by relatively long scanning time, which is required for the motion of the transducer over the entire scanning area. This time may be reduced by using a multi-channel acoustical system which has several identical transducers arranged as an array and is mounted on a mechanical scanner so that each transducer scans only a fraction of the total area. The resulting image is formed as a combination of all acquired partial data sets. The mechanical instability of the scanner, as well as the difference in parameters of the individual transducers causes a misalignment of the image fractures. This distortion may be partially compensated for by the introduction of constant or dynamical signal leveling and data shift procedures. However, a reduction of the random instability component requires more advanced algorithms, including auto-adjustment of processing parameters. The described procedure was implemented into the prototype of an ultrasonic fingerprint reading system. The specialized cylindrical scanner provides a helical spiral lens trajectory which eliminates repeatable acceleration, reduces vibration and allows constant data flow on maximal rate. It is equipped with an array of four spherically focused 50 MHz acoustic lenses operating in pulse-echo mode. Each transducer is connected to a separate channel including pulser, receiver and digitizer. The output 3D data volume contains interlaced B-scans coming from each channel. Afterward, data processing includes pre-determined procedures of constant layer shift in order to compensate for the transducer displacement, phase shift and amplitude leveling for compensation of variation in transducer characteristics. Analysis of statistical parameters of individual scans allows adaptive eliminating of the axial misalignment and mechanical vibrations. Further 2D correlation of overlapping partial C-scans will realize an interpolative adjustment which essentially improves the output image. Implementation of this adaptive algorithm into a data processing sequence allows us to significantly reduce misreading due to hardware noise and finger motion during scanning. The system provides a high quality acoustic image of the fingerprint including different levels of information: fingerprint pattern, sweat porous locations, internal dermis structures. These additional features can effectively facilitate fingerprint based identification. The developed principles and algorithm implementations allow improved quality, stability and reliability of acoustical data obtained with the mechanical scanner, accommodating several transducers. General principles developed during this work can be applied to other configurations of advanced ultrasonic systems designed for various biomedical and NDE applications. The data processing algorithm, developed for a specific biometric task, can be adapted for the compensation of mechanical imperfections of the other devices.