From 6d7acd11eea43817a2e139bb8359f2c7de75224b Mon Sep 17 00:00:00 2001
From: CroneKorkN <i@ckn.li>
Date: Sun, 1 Jun 2025 14:14:00 +0200
Subject: [PATCH] wip

---
 process | 25 +++++++++++++++++--------
 1 file changed, 17 insertions(+), 8 deletions(-)

diff --git a/process b/process
index 215e909..9a1e5a9 100755
--- a/process
+++ b/process
@@ -21,6 +21,7 @@ ADJACENCY_FACTOR = 2 # area to look for noise around the target frequency
 AMPLITUDE_THRESHOLD = 200 # relative DB (rDB) (because not calibrated)
 BLOCK_SECONDS = 3 # seconds (longer means more frequency resolution, but less time resolution)
 DETECTION_DISTANCE = 30 # seconds (minimum time between detections)
+BLOCK_OVERLAP = 0.8 # overlap between blocks (0.8 means 80% overlap)
 
 
 def process_recording(filename):
@@ -30,23 +31,31 @@ def process_recording(filename):
     date_string_from_filename = os.path.splitext(filename)[0]
     recording_date = datetime.datetime.strptime(date_string_from_filename, "%Y-%m-%d_%H-%M-%S.%f%z")
 
-    # get samplerate and blocksize
+    # get data and metadata from recording
     path = os.path.join(RECORDINGS_DIR, filename)
-    info = soundfile.info(path)
-    samplerate = info.samplerate
-    block_samples = int(BLOCK_SECONDS * samplerate)
+    sound, samplerate = soundfile.read(path)
+    samples_per_block = int(BLOCK_SECONDS * samplerate)
+    overlap = int(samples_per_block * BLOCK_OVERLAP)
+
+    # get median amplitude for normalization
+    complex_amplitudes_global = rfft(sound)
+    median_amplitude = np.median(np.abs(complex_amplitudes_global))
+    print(f'median amplitude: {median_amplitude:.5f}rDB')
 
     # initialize to a very old date
     last_detection_at = datetime.datetime.min.replace(tzinfo=recording_date.tzinfo)
     is_detecting = False
 
-    # iterate blocks
-    for num, block in enumerate(soundfile.blocks(path, blocksize=block_samples, overlap=int(block_samples*0.8))):
+    # read blocks of audio data with overlap from sound variable
+    for num in range(0, len(sound) // (samples_per_block - overlap)):
+        start = num * (samples_per_block - overlap)
+        end = start + samples_per_block
+        block = sound[start:end]
+
         block_date = recording_date + datetime.timedelta(seconds=num * BLOCK_SECONDS)
         labels = rfftfreq(len(block), d=1/samplerate)
         complex_amplitudes = rfft(block)
         absolute_amplitudes = np.abs(complex_amplitudes)
-        median_amplitude = np.median(absolute_amplitudes)
         amplitudes = absolute_amplitudes / median_amplitude
 
         # get amplitudes only between 100 and 1000 Hz
@@ -79,7 +88,7 @@ def process_recording(filename):
                 last_detection_at = block_date
                 print("🔊")
 
-            print(f'{block_date}: {max_amplitude:.1f}rDB @ {max_freq:.1f}Hz (noise {noise:.3f}rDB, median {median_amplitude:.3f}rDB)')
+            print(f'{block_date}: {max_amplitude:.1f}rDB @ {max_freq:.1f}Hz (noise {noise:.3f}rDB)')
         else:
             is_detecting = False