Raven Batch-detector error

Question

I'm trying to analyze a large set of data so I'm using a band limited energy batch-detector. I was running individual detection (Tools: Detector) to fine tune my detector parameters. When running a batch detection with the same parameters as the individual detector on the same files, I receive completely different outputs. I reran the single with diagnostics on and the values I'm getting in the batch detection aren't even in the table of "Too Long, Too short, Too sparse."

I have double checked all my parameters and they're all the same. I've gone back through by hand and checked sample sections and there is no overlap in the detector and the batch detector. I even uninstalled and reinstalled Raven Pro.

I'm using 1.6.3 - if anyone has any experience with this or ideas please let me know

Answer 1

there are a number of possible explanations for your observations. I will list a few below, but would encourage you to request support by emailing [email protected] to establish the particulars of your case. I can promise a timely response.

1. Default spectrogram parameters: When you run the Band-Limited Energy Detector (BLED) in batch mode, the default window size and DFT size are 256 samples. In contrast, when you run BLED in interactive or full mode, the default window size and DFT size are 512 samples (assuming you have not set up a custom default window preset). This difference can lead to different results.

2. Noise Block Progression: The first stage of the BLED estimates a series of noise levels that are used to calculate SNR to apply to the SNR Treshold (which you set in the Noise tab in the Configure Band Limited Energy Detector dialog). Noise level is estimated in a series of sound blocks with duration set by the Block Size parameter in the Noise tab. This noise block is progressed through the recording by the increment set in the Hop Size parameter in the Noise tab. When you run BLED in full mode, there is a single, uninterrupted progression of noise blocks through the appended sound samples of all sound files open in the sound window. In contrast, when you run BLED in batch mode the progression of noise blocks starts at the beginning of each sound file added to the Batch Detect dialog. This difference in which noise blocks are used to estimate noise levels can result in different results for batch detection compared to full detection.

3. Short Sound Files: Sound files with length less that the Block Size will not have any detections in them if you run the BLED in a batch mode. However, when you run BLED in interactive or full mode there may well be detections found in such short sound files because sound samples from the preceding or succeeding sound files can be added to the noise block.

4. Timekeeping: When you open multiple sound files into a sound window in Raven, sound samples from those sound files are appended. The time position of selections in the selection time are relative to the beginning of the first sound file.

In contrast, when you set up a batch detection by adding individual sound files to the Batch Detect dialog, the output will consist of one selection table for each sound file, with time positions reported relative to the beginning of single, unappended sound files. If you open these selection tables into a sound window with multiple sound files, the selections will appear at the wrong time.

I hope that helps!

By the way, Raven Pro 1.6.4 has been available since July 2022. I would encourage you to email [email protected] to discuss why automatic updates aren't working on your computer.