Decoding Temporal Features of Birdsong Through Neural Activity Analysis

<!-- ✨ Dataset Description for
“Decoding Temporal Features of Birdsong Through Neural Activity Analysis” (Ahmadi et al., 2025) -->
<h2>Overview</h2>
<p>
This repository contains every dataset, analysis output and visualisation that
support the manuscript <strong>“Decoding Temporal Features of Birdsong Through Neural
Activity Analysis”</strong> by <em>Amirmasoud Ahmadi, Hermina Robotka, Manfred Gahr
and Frederic Theunissen (2025)</em>. Neural activity was recorded in the auditory
pallium of adult zebra finches while they listened to unfamiliar conspecific songs.
All archives are provided as <code>.mat</code>, <code>.csv</code> or
<code>.avi</code> files to maximise cross-platform usability.
</p>

<h2>1 · Decoding Results</h2>
<ul>
<li><code>LFP_Decoding_Results.zip</code>: predictions of <em>Events</em>,
<em>Envelopes</em> and <em>Landmarks</em> from local-field potentials.</li>
<li><code>MUAe_Decoding_Results.zip</code>: identical analyses on multi-unit
activity envelopes.</li>
<li><code>LFP + MUAe_Decoding_Results.zip</code>: performance obtained when
LFP and MUAe feature vectors are concatenated.</li>
</ul>

<h2>2 · Single-Unit Responses</h2>
<p>
<code>Single_Unit_Response_To_Song_Playback.zip</code> supplies spike trains and
peri-stimulus time histograms for <strong>423</strong> well-isolated neurons,
enabling cell-by-cell comparisons with the population-based decoders.
</p>

<!-- 🔥 THE MOST IMPORTANT RESULTS AT A GLANCE 🔥 -->
<h2><strong>3 · SUMMARY TABLES (KEY DATASET)</strong></h2>
<p>
<code>Summary_Results_Table.zip</code> compiles the headline decoding statistics in
three clearly labelled folders:
</p>
<ul>
<li><strong>LFP/</strong>
<ul>
<li><code>LFP_EventDetection.csv</code></li>
<li><code>LFP_Env.csv</code></li>
<li><code>LFP_EnvelopeLandmarks.csv</code></li>
</ul>
</li>
<li><strong>MUAe/</strong>
<ul>
<li><code>MUAe_EventDetection.csv</code></li>
<li><code>MUAe_Env.csv</code></li>
<li><code>MUAe_EnvelopeLandmarks.csv</code></li>
</ul>
</li>
<li><strong>Fusion (LFP_MUAe)/</strong>
<ul>
<li><code>Fusion_EventDetection.csv</code></li>
<li><code>Fusion_Env.csv</code></li>
<li><code>Fusion_EnvelopeLandmarks.csv</code></li>
</ul>
</li>
</ul>
<p>
Each file reports overall accuracy, Cohen kappa, syllable-level and silent-period
accuracies, together with full metadata (<code>Birds_Name</code>,
<code>Sex_Birds</code>, <code>Song_Number</code>, <code>Depth_Record</code>, etc.).
These metrics reproduce the numbers in Table 1 of the manuscript.
</p>

<h2>4 · Figure Source Data</h2>
<p>
Six archives (<code>Figure2_Data.zip</code>, <code>Figure3_Data.zip</code>,
<code>Figure4_Data.zip</code>, <code>Figure5_Data.zip</code>,
<code>Figure6_Data.zip</code>, <code>Figure7_Data.zip</code>) recreate every panel of
the main figures. <code>SupFig_Data.zip</code> holds all supplementary figure data.
Each archive contains MATLAB matrices and comma-separated tables.
</p>

<h2>5 · Supplementary Video</h2>
<p>
<code>Figure4_3D_Video.zip</code> contains an <code>.avi</code> file showing a
rotating three-dimensional map of decoding accuracy across recording sites,
corresponding to Figure 4 of the paper.
Additional demonstration videos related to the study can be found on YouTube at
<a href="https://www.youtube.com/@Amir_Channel_Sci"
target="_blank" rel="noopener">www.youtube.com/@Amir_Channel_Sci</a>.
</p>

<h2>6 · Code Availability</h2>
<p>
All scripts that generate the manuscript figures and the core routines used for
neural-signal processing are openly available at
<a href="https://github.com/amirmasoud92/ZF_Neural_Decoding"
target="_blank" rel="noopener">https://github.com/amirmasoud92/ZF_Neural_Decoding</a>.
</p>

<p><em>Please cite both the manuscript and this dataset if you reuse any of these files.</em></p>