findSimilarity

バージョン 1.0 (2.71 KB) 作成者: Matt Flax
A faster implementation of findsignal
ダウンロード: 17
更新 2021/10/20

findSimilarity

Finding a similar signal in a longer signal is a common problem in signal processing. Waveform Similarity Overlap Add (WSOLA) is an example of an algorithm which uses signal similarity to speed up and slow down audio without changing its pitch.

The brute force method to find a similar signal between a reference (s) and a longer waveform (y) is based on RMS value of the distance :

Ns=size(s,1);
res=buffer(y,Ns,Ns-1,'nodelay');
res=s-res;
errM=rms(res);
[err,nI]=min(errM);

This brute force method is slow because it uses a huge amount of memory in buffering the signal with one sample overlap and the computational complexity of the RMS function.

Methods based on the DFT (for example Mueen's implementation) break the nonlinear distance measure into a linear component which is convolved using the DFT :

(s-y)^2 = s^2 - 2s*y + y^2

The s*y convolution is performed in the DFT domain which is a speed up for large vectors. However for smaller s and y vectors the computation is still not efficient enough.

The solution proposed here reduces the order of the computation by breaking it into two steps. The first step computes the distance measure on a smaller signal space looking for the possible location of the global minimum. The second iteration does a complete search of the global minimum's region looking for the exact result.

引用

Matt Flax (2024). findSimilarity (https://github.com/flatmax/findSimilarity/releases/tag/v1.0), GitHub. 取得済み .

MATLAB リリースの互換性
作成: R2021b
すべてのリリースと互換性あり
プラットフォームの互換性
Windows macOS Linux
タグ タグを追加

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!
バージョン 公開済み リリース ノート
1.0

この GitHub アドオンでの問題を表示または報告するには、GitHub リポジトリにアクセスしてください。
この GitHub アドオンでの問題を表示または報告するには、GitHub リポジトリにアクセスしてください。