Versions Compared

Old Version 2

changes.mady.by.user Alena Kniazeva

Saved on

compared with

New Version Current

changes.mady.by.user Alena Kniazeva

Saved on

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.

Hardware and audio data requirements for the operation of the Audio classification IV are given in the table:

Microphone
The Microphone object must be enabled (see Microphone)

The microphone must be physically connected, correctly recognized by the operating system as an audio capture device, and enabled in Axxon One (see Microphone). Depending on the microphone type (analog, USB, or IP microphone), you must ensure a stable power supply and adequate signal levels

StreamThe number of channels is Mono (single-channel stream). Stereo or multi-channel streams must be downmixed to mono before being fed to the detector
StreamThe detector requires a single-channel (mono) audio stream

Audio data format

Audio data must be transferred in
The data format is 16-bit signed integer PCM
format
(little-endian). Other formats (32-bit float, A-law, μ-law, and so on) require conversion on the source side
Sample rateSupported
sample rate:
sampling rates are 8000, 16000, 32000, 48000 Hz.
We recommend 16000 HzHardware

For the correct operation of the audio analytics algorithms, you must disable all embedded DSP features of audio processing on your audio hardware or in drivers, as they can distort the original audio signal and interfere with recognition. These features include:

  • Automatic Gain Control (AGC)
  • Noise Suppression/Cancellation
  • Beamforming
  • Echo Cancellation
  • Equalizers (EQ) and High Pass Filters (HPF)
The recommended sampling rate is 16000 Hz. Lower rates reduce recognition quality; higher rates are redundant and increase the load on transmission and processing channels
Programmable gain amplifier

The device must be equipped with a programmable gain amplifier (PGA) with small gain adjustment steps (1 dB) and a low self-noise. This allows for precise adjustment of the microphone's sensitivity without introducing additional nonlinear distortion

Digital signal processing (DSP)

All audio processing algorithms that could distort the original signal and interfere with recognition must be disabled on the hardware and drivers:

  • Automatic gain control (AGC)
  • Noise suppression/cancellation
  • Beamforming
  • Acoustic echo cancellation (AEC)
  • Equalizers (EQ) and high-pass filters (HPF)
  • De‑esser
  • Spatial sound, virtual surround

Each feature must have an embedded disable method (physical switch, driver settings, control panel). Embedded audio processing devices interfere with the operation of analytical algorithms

Sound pressure level (SPL)
For the detector to generate an event, the

The sound pressure level

(SPL) from

of the target event

must be within the following ranges (measured

, measured at the point of microphone installation with a sound level meter (Class 2 according to IEC 61672), must meet the following operating conditions:

  • In a noisy

environment: at least 80–82 dB

  • environment—80–82 dB or higher (production facility, busy street, room with operating equipment).

  • In a quiet environment—58–60 dB or higher (office, meeting room, living room).

 If the signal level does not meet the requirements, the system can fail to detect events or generate false alarms

Operating system

  • Windows OS:

    • Use the standard audio API, and redirect the stream via WASAPI or ASIO.

    • In the audio control panel, go to Device Properties → Additional device properties → Enhancements → clear the Disable all enhancements checkbox. If the tab is missing, check the settings in the utility of your sound card driver.

  • Linux OS:
    • Use ALSA-compatible drivers with basic support. When you use PulseAudio or PipeWire, ensure they are configured for "transparent" mode (flat volumes, noise reduction disabled, and attenuation disabled).
    • When you work with embedded codecs, refer to documented Device Tree Source (DTS) examples to ensure correct audio data transmission without additional processing
In a quiet environment: at least 58–60 dB