Clean spectral analysis

birdnet-pi/rootfs/helpers/spectral_analysis.py

@@ -1,70 +0,0 @@
-import numpy as np
-import scipy.io.wavfile as wavfile
-import matplotlib.pyplot as plt
-import os
-import glob
-import sys  # Import the sys module
-
-from utils.helpers import get_settings
-
-# Dependencies /usr/bin/pip install numpy scipy matplotlib
-
-# Define the directory containing the WAV files
-conf = get_settings()
-input_directory = os.path.join(conf['RECS_DIR'], 'StreamData')
-output_directory = os.path.join(conf['RECS_DIR'], 'Extracted/Charts')
-
-# Ensure the output directory exists
-if not os.path.exists(output_directory):
-    os.makedirs(output_directory)
-
-# Check if a command-line argument is provided
-if len(sys.argv) > 1:
-    # If an argument is provided, use it as the file to analyze
-    wav_files = [sys.argv[1]]
-else:
-    # If no argument is provided, analyze all WAV files in the directory
-    wav_files = glob.glob(os.path.join(input_directory, '*.wav'))
-
-# Process each file
-for file_path in wav_files:
-    # Load the WAV file
-    sample_rate, audio_data = wavfile.read(file_path)
-
-    # If stereo, select only one channel
-    if len(audio_data.shape) > 1:
-        audio_data = audio_data[:, 0]
-
-    # Apply the Hamming window to the audio data
-    hamming_window = np.hamming(len(audio_data))
-    windowed_data = audio_data * hamming_window
-
-    # Compute the FFT of the windowed audio data
-    audio_fft = np.fft.fft(windowed_data)
-    audio_fft = np.abs(audio_fft)
-
-    # Compute the frequencies associated with the FFT values
-    frequencies = np.fft.fftfreq(len(windowed_data), d=1/sample_rate)
-
-    # Select the range of interest
-    idx = np.where((frequencies >= 150) & (frequencies <= 15000))
-
-    # Calculate the saturation threshold based on the bit depth
-    bit_depth = audio_data.dtype.itemsize * 8
-    max_amplitude = 2**(bit_depth - 1) - 1
-    saturation_threshold = 0.8 * max_amplitude
-
-    # Plot the spectrum with a logarithmic Y-axis
-    plt.figure(figsize=(10, 6))
-    plt.semilogy(frequencies[idx], audio_fft[idx], label='Spectrum')
-    plt.axhline(y=saturation_threshold, color='r', linestyle='--', label='Saturation Threshold')
-    plt.xlabel("Frequency (Hz)")
-    plt.ylabel("Amplitude (Logarithmic)")
-    plt.title(f"Frequency Spectrum (150 - 15000 Hz) - {os.path.basename(file_path)}")
-    plt.legend()
-    plt.grid(True)
-
-    # Save the plot as a PNG file
-    output_filename = os.path.basename(file_path).replace('.wav', '_spectrum.png')
-    plt.savefig(os.path.join(output_directory, output_filename))
-    plt.close()  # Close the figure to free memory

birdnet-pi/rootfs/helpers/spectral_analysis.sh

@@ -1,65 +0,0 @@
-#!/usr/bin/env bash
-# Performs the recording from the specified RTSP stream or soundcard
-set +u
-# shellcheck disable=SC1091
-source /etc/birdnet/birdnet.conf
-
-# Read the logging level from the configuration option
-# shellcheck disable=SC2154
-LOGGING_LEVEL="${LogLevel_BirdnetRecordingService:-error}"
-# If empty for some reason default to log level of error
-[ -z "$LOGGING_LEVEL" ] && LOGGING_LEVEL='error'
-# Additionally if we're at debug or info level then allow printing of script commands and variables
-if [ "$LOGGING_LEVEL" == "info" ] || [ "$LOGGING_LEVEL" == "debug" ];then
-  # Enable printing of commands/variables etc to terminal for debugging
-  set -x
-fi
-
-[ -z "$RECORDING_LENGTH" ] && RECORDING_LENGTH=15
-[ -d "$RECS_DIR"/StreamData ] || mkdir -p "$RECS_DIR"/StreamData
-
-filename="Spectrum_$(date "+%Y-%m-%d_%H:%M").wav"
-
-if [ -n "${RTSP_STREAM:-}" ];then
-  # Explode the RSPT steam setting into an array so we can count the number we have
-  RTSP_STREAMS_EXPLODED_ARRAY=("${RTSP_STREAM//,/ }")
-
-  while true;do
-
-    # Initially start the count off at 1 - our very first stream
-    RTSP_STREAMS_STARTED_COUNT=1
-    FFMPEG_PARAMS=""
-
-    # Loop over the streams
-    for i in "${RTSP_STREAMS_EXPLODED_ARRAY[@]}"
-    do
-      # Map id used to map input to output (first stream being 0), this is 0 based in ffmpeg so decrement our counter (which is more human readable) by 1
-      MAP_ID="$((RTSP_STREAMS_STARTED_COUNT-1))"
-      # Build up the parameters to process the RSTP stream, including mapping for the output
-      FFMPEG_PARAMS+="-vn -thread_queue_size 512 -i ${i} -map ${MAP_ID}:a:0 -t ${RECORDING_LENGTH} -acodec pcm_s16le -ac 2 -ar 48000 file:${RECS_DIR}/StreamData/$filename "
-      # Increment counter
-      ((RTSP_STREAMS_STARTED_COUNT += 1))
-    done
-
-  # Make sure were passing something valid to ffmpeg, ffmpeg will run interactive and control our loop by waiting ${RECORDING_LENGTH} between loops because it will stop once that much has been recorded
-  if [ -n "$FFMPEG_PARAMS" ];then
-    ffmpeg -hide_banner -loglevel "$LOGGING_LEVEL" -nostdin "$FFMPEG_PARAMS"
-  fi
-
-  done
-else
-  if pgrep arecord &> /dev/null ;then
-    echo "Recording"
-  else
-    if [ -z "${REC_CARD}" ];then
-      arecord -f S16_LE -c"${CHANNELS}" -r48000 -t wav --max-file-time "${RECORDING_LENGTH}"\
-	      	      	       --use-strftime "${RECS_DIR}"/StreamData/"$filename"
-    else
-      arecord -f S16_LE -c"${CHANNELS}" -r48000 -t wav --max-file-time "${RECORDING_LENGTH}"\
-        -D "${REC_CARD}" --use-strftime "${RECS_DIR}"/StreamData/"$filename"
-    fi
-  fi
-fi
-
-# Create the spectral analysis
-"$PYTHON_VIRTUAL_ENV" "$HOME"/BirdNET-Pi/scripts/spectral_analysis.py