GUI-based MATLAB script for robust segmentation of rodents' ultrasonic vocalizations

• All stuffs are migrating to a GitHub repository (not completed)

• Brief presentation for USVSEG (in Japanese) in SlideShare

• Download software ...

  • latest version (version 0.9, revision 2) from here

    • changes in CSV output: duration is now in milliseconds but not seconds; additionally output "meanfreq" and "cvfreq". (See change history)

  • stable previous version (version 0.8, revision 7) from here

• User Guide documents: English (ver.211103) / Japanese (ver.211103)

• Please check [this site] for additional information, made by our collaborator, Dr. Kouta Kanno.

• A paper reporting segmentation performances has been published! (Tachibana et al., 2020, PLoS ONE)

• Dataset (total ~1GB) used for performance tests in the paper is available on Zenodo repository.

• A part of mice USV in the dataset was obtained from Data Sharing Platform of "Individuality" project.

Citation: Tachibana RO, Kanno K, Okabe S, Kobayasi KI, Okanoya K (2020) USVSEG: A robust method for segmentation of ultrasonic vocalizations in rodents. PLoS ONE 15(2): e0228907.

Weight-reduced FinchScope

• "FinchScope" is a 3D-printing DIY version of the miniscope (head-mount fluorescent microscope for one-photon calcium imaging), which has been originally developed in Tim Gardner lab (https://github.com/gardner-lab/FinchScope).

• I'm trying to reduce the weight of FinchScope chassis for direct implantation onto the bird's head. Currently the scope is ~1.7 gram with all parts (but excluding tethering cable). Please email to me if you want to test it.

Sync Pulse Interface

• One of most important issues for the vocalization researches is the reliability on timestamp of acoustic data.

• For a tight synchronization between audio inputs and calcium event movies, I designed a pulse generator that receives a video signal (NTSC), and provides ~1-second interval pulse in synchronization of the video frame change. This generator also inserts a white dot in the video frame at the timing of each pulse, which allows us to synchronize tightly the audio and video signal in offline analysis. Ask me if you want to test.

• The interface board is designed as a "shield" for Arduino UNO, which is based on the Video Experimenter shield

• Semi-automatic detection and classification of birdsong syllables by a simple machine-learning technique: Tachibana et al (2014) PLOE ONE

  • This method trains a linear SVM by manually labeled one-minuet excerpts, then generalizes the syllable categories to remained one-day recording data.

  • Syllable labeling is one of the most laborious parts of songbird research for lazier scholar like me... So, I've developed this method to spoil myself :)

  • You can find the script for making the acoustical feature set here.

in prep