Step 3: NAS processing architecture
Step 3 is the most important OpenNAS wizard step because the user has to select the NAS architecture. This is, the filters' architecture to perform the audio signal decomposition into specific frequency signals. Filters' connection can be established either in cascade or in parallel. By default, cascade architecture is selected since it is biologically inspired on how the cochlea works. But new architectures, as the parallel one, are rising up on the new designs.
The user can choose between three options:
Next, each architecture's features are listed and explained in detail.
Order of the Spiking Low-Pass Filter.
- Order 2.
This parameter can not be changed by the user. Its value is shown just to remember to the user the number of frequency channels selected in step 2.
- From 2 to 1024.
The start frequency in which the audio input signal is going to decompose. This parameter establishes the middle frequency of the first SBPF. By default, this parameter is set to 20 Hz, where human audible sounds range starts approximately. But it could start at 1 Hz.
- 20 Hz.
The stop frequency in which the audio input signal is going to decompose. This parameter establishes the middle frequency of the last SBPF. By default, this parameter is set to 22 KHz, where human audible sound range ends approximately, although it could finish as higher as you need. However, you must consider the Start Freq. value to set the Stop Freq value.
- 22 KHz.
By changing this parameter, the number of output spikes from the spike-based band-pass filter can be controlled. Note that this attenuation value is shared by the filters bank. Then, each SBPF will have the same output attenuation. By default, the output attenuation value is -12 (in dB). For speech recognition, we recommend to decrease the gain by setting this parameter to -36.
- -12.
On this table, a summary of the middle frequency of each SBPF filter is shown. Each middle frequency has been calculated by taking into account the Start frequency, the Stop frequency and the number of frequency channels.
On this table, a summary of the cut-off frequency of each SLPF filter is shown. Each cut-off frequency has been calculated by taking into account the Start frequency, the Stop frequency and the number of frequency channels, such that those frequencies make possible to obtain the desired band-pass filter's middle frequency.
On this table, a summary of the output attenuation of each SBPF filter is shown. Each middle frequency has been calculated by taking into account the Start frequency, the Stop frequency and the number of frequency channels.
Order of the Spiking Low-Pass Filter.
- Order 2.
This parameter can not be changed by the user. Its value is shown just to remember to the user the number of frequency channels selected in step 2.
- From 2 to 1024.
The start frequency in which the audio input signal is going to decompose. This parameter establishes the middle frequency of the first SBPF. By default, this parameter is set to 20 Hz, where human audible sounds range starts approximately. But it could start at 1 Hz.
- 20 Hz.
The stop frequency in which the audio input signal is going to decompose. This parameter establishes the middle frequency of the last SBPF. By default, this parameter is set to 22 KHz, where human audible sound range ends approximately, although it could finish as higher as you need. However, you must consider the Start Freq. value to set the Stop Freq value.
- 22 KHz.
By changing this parameter, the number of output spikes from the spike-based band-pass filter can be controlled. Note that this attenuation value is shared by the filters bank. Then, each SBPF will have the same output attenuation. By default, the output attenuation value is -12 (in dB). For speech recognition, we recommend to decrease the gain by setting this parameter to -36.
- -12.
On this table, a summary of the middle frequency of each SBPF filter is shown. Each middle frequency has been calculated by taking into account the Start frequency, the Stop frequency and the number of frequency channels.
On this table, a summary of the cut-off frequency of each SLPF filter is shown. Each cut-off frequency has been calculated by taking into account the Start frequency, the Stop frequency and the number of frequency channels, such that those frequencies make possible to obtain the desired band-pass filter's middle frequency.
On this table, a summary of the output attenuation of each SBPF filter is shown. Each middle frequency has been calculated by taking into account the Start frequency, the Stop frequency and the number of frequency channels.
- Order 2
This parameter can not be changed by the user. Its value is shown just to remember to the user the number of frequency channels selected in step 2.
- From 2 to 1024.
The start frequency in which the audio input signal is going to decompose. This parameter establishes the middle frequency of the first SBPF. By default, this parameter is set to 20 Hz, where human audible sounds range starts approximately. But it could start at 1 Hz.
- 20 Hz.
The stop frequency in which the audio input signal is going to decompose. This parameter establishes the middle frequency of the last SBPF. By default, this parameter is set to 22 KHz, where human audible sound range ends approximately, although it could finish as higher as you need. However, you must consider the Start Freq. value to set the Stop Freq value.
- 22 KHz.
Filter Q factor.
By changing this parameter, the number of output spikes from the spike-based band-pass filter can be controlled. Note that this attenuation value is shared by the filters bank. Then, each SBPF will have the same output attenuation. By default, the output attenuation value is -12 (in dB). For speech recognition, we recommend to decrease the gain by setting this parameter to -36.
- -12.
On this table, a summary of the middle frequency of each SBPF filter is shown. Each middle frequency has been calculated by taking into account the Start frequency, the Stop frequency and the number of frequency channels.
On this table, a summary of the filters' Q factor is shown.
On this table, a summary of the output attenuation of each SBPF filter is shown. Each middle frequency has been calculated by taking into account the Start frequency, the Stop frequency and the number of frequency channels.
