Local Gradients for Pitch Resistant Audio Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Audio matching systems face challenges in identifying pitch-shifted audio samples due to alterations in interest points, leading to difficulties in matching distorted signals, particularly in cases where small pitch shifts occur.
Innovation Solution
The system generates gradients related to interest points within descriptors, including horizontal, vertical, and diagonal gradients, which characterize the spectrogram neighborhood by expressing energy ratios between regions, making the system more robust to pitch shift distortion.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If pitch shifting is applied to audio samples, then the frequency of interest points is shifted, but this causes difficulty in matching the pitch shifted signal
Solution Approach 1:
The patent segments the audio signal analysis into distinct components: interest point detection, gradient computation, and descriptor generation. By dividing the spectrogram into local neighborhoods around interest points and computing gradients separately for each, the system can maintain precision while adapting to pitch shifts. Each segment (local neighborhood) is processed independently to capture local structural information that remains invariant to global pitch shifting.
Solution Approach 2:
The patent introduces gradient information as an additional dimension to the traditional interest point descriptors. Instead of relying solely on interest point locations and frequencies, the system computes gradient magnitudes and directions in the time-frequency plane, adding spatial derivative information that helps distinguish pitch-shifted signals. This dimensional expansion allows the system to maintain matching accuracy despite frequency shifts.
2Reliability
If small pitch shifts occur in audio samples, then the pitch shift is hard to notice for listeners, but it presents difficult challenges in matching the pitch shifted signal
Solution Approach 1:
The patent applies local quality by computing gradients in local neighborhoods around each interest point rather than analyzing the entire spectrogram globally. This local analysis captures subtle structural changes caused by small pitch shifts that might be imperceptible to listeners but are detectable through gradient computations. Each local region's gradient properties provide reliable matching features even when global pitch shifts are minimal.
Solution Approach 2:
The system performs preliminary gradient computation on the spectrogram before conducting the actual audio matching. By pre-calculating gradient magnitudes and directions for all interest points, the system prepares pitch-shift-invariant features in advance, making the subsequent matching process more reliable even when small pitch shifts occur that are difficult to detect through traditional methods.
Data Source
AI summary
System and methods for characterizing interest points within a descriptor are disclosed herein. The systems include generating a set of interest points related to an audio sample. A set of gradients relating to respective interest points in the set of interest points can be generated. A set of descriptors can then be generated based upon the set of interest points and the set of gradients and used in comparison to reference descriptors to identify the audio sample. The disclosed systems and methods provide for an audio matching system robust to pitch-shift distortion by using gradients that characterize the time-frequency neighborhood around an interest point rather than solely relying on interest points themselves. Thus, the disclosed system and methods result in more accurate audio identification.