atools: A set of audio tools

1. atools-findecho.py

This tool will play a short sample and record simultanously looking for the first duplication of the played audio. Typical use case is to record roundtrip delay on a single device or end to end delay in a communication system between devices. The audio directory contains a 8kHz chirop which can be used as a signal.

$ python/atools-findecho.py --help
usage: atools-findecho.py [-h] -i IMPULSE -o OUTPUT_CSV [-t SECONDS] [-d DELAY]
                   [-s SAVE] [-v]

optional arguments:
  -h, --help            show this help message and exit
  -i IMPULSE, --impulse IMPULSE
  -o OUTPUT_CSV, --output_csv OUTPUT_CSV
  -t SECONDS, --seconds SECONDS
  -d DELAY, --delay DELAY
  -s SAVE, --save SAVE
  -v, --verbose

Example:

$ python/atools-findecho.py -i audio/chirp_300-3k.wav -o pixel3 -t 30

2. atools-multitone.py

This tool generates multitone wav files. A multitone is a series of (synchronized) sin signals at consecutive octaves, followed by silence. Tool allows selecting the first and last frequencies, the scale (to get louder/softer signals), the total duration, and the duration of the active part (the multitone concatenates periods of sin signal followed by silences).

Example:

$ python/atools-multitone.py --duration-sec 10 --active-ratio 0.1 \
    multitone.duration_sec_10.active_ratio_0_1.16000_hz.wav

Figure 1 shows a (audacity) spectogram of the output file of the last command (available here).

Usage:

$ ./atools-multitone.py --help
usage: atools-multitone.py [-h] [-d] [--quiet] [--samplerate SAMPLERATE]
                    [--duration-sec DURATION_SEC]
                    [--active-ratio ACTIVE-RATIO] [--f0 F0] [--f1 F1]
                    [--scale SCALE] [--signal [multitone]]
                    output-file

positional arguments:
  output-file           output file

options:
  -h, --help            show this help message and exit
  -d, --debug           Increase verbosity (use multiple times for more)
  --quiet               Zero verbosity
  --samplerate SAMPLERATE
                        use SAMPLERATE samplerate
  --duration-sec DURATION_SEC
                        use DURATION_SEC duration_sec
  --active-ratio ACTIVE-RATIO
                        use active_ratio
  --f0 F0               use F0 for start chirp frequency
  --f1 F1               use F1 for end chirp frequency
  --scale SCALE         use scale
  --signal [multitone]  signal arg

3. feedbackcount.py

This tool uses the method of measuring distances between feedback signals. One of the issues with the findecho.py method is that sometimes feedback is hard to get a round. With a realtive short distance between the microphone and the speaker and/or relfective surfaces, frequencies from the speaker signal are being picket up and amplified by the microphone. If there is a delay in the system this will lead to a pulsating system and the idea here is that the frequency of the pulses is directly correlated to the delay time. The system will do a recording (with or without providing a stimuli to start the feedback) and analyse the spectrum, pick out a strong signal and bandpass filter around it. That signal is then filtered further and the the peaks are identified and the distance between them is counted and put in a histogram where the highest bin will determine the delay.

One of the complexities with the feedback analysis is that it is possible that a multitude of signals are feeding simultanously and the script needs to separate them. The idea is that all related peaks which are a function of the real delay will have the same periodicity and the histogram will peak for that period. To measure it is probably best to have a system that does not feedback by itself but needs a stimuli to start the loop. Also, limiting the number of stimuli signals makes is easier to not have them confusing the measurements.

usage: atools-feedbackcount.py [-h] [-i IMPULSE] [--source SOURCE] -o OUTPUT_CSV [-t SECONDS] [-d DELAY] [-s SAVE] [-v] [-mi MINTIMEMS] [-mx MAXTIMEMS] [-nf]

optional arguments: -h, --help show this help message and exit -i IMPULSE, --impulse IMPULSE --source SOURCE If set, calculation will be done on this file -o OUTPUT_CSV, --output_csv OUTPUT_CSV -t SECONDS, --seconds SECONDS -d DELAY, --delay DELAY -s SAVE, --save SAVE Save and concatenate the recorded audio -v, --verbose -mi MINTIMEMS, --mintimems MINTIMEMS Shortes pulse to consider in feedback analysis -mx MAXTIMEMS, --maxtimems MAXTIMEMS Shortes pulse to consider in feedback analysis -nf, --nofilter Do not filter for frequency peaks

Example:

$ python/atools-feedbackcount.py -i audio/chirp_300-3k.wav -o pixel3 -t 30 -d 10

The command above will start a measurement using the '-i' as stimuli running for 30 s and the stimuli to be player every 10th seconds.

• The bandpass filter can be skipped and limits on the shortest/longest delay to look for can be set. In noisy environment slow periodict signals like air condition can easily be misstaken for the real signal. To see what signal is analyzed, use the vebose option.
• Using the '--source' a prerecorded signal containing a feedback is analyzed.
• If measuring a system already in feedback condition skip the impulse and just record.