SBAA600 October 2024 TAA5212 , TAC5111 , TAC5112 , TAC5211 , TAC5212
Initial Learning Period (ILP) is the amount of time the VAD algorithm takes to adjust to the background noise environment, from the time instant VAD is turned on. In practical applications, the device is soaked in the background noise by default and the VAD interrupt can be low during the ILP followed by which the detector can start the voice detection. After the Initial Learning Period is completed, the device calculates the coefficients required for the algorithm and the VAD starts to work as expected within the designed for accuracy limit. The device can continue to calculate coefficients after the ILP is over and improve accuracy as long as the VAD remains powered up in a given environment. The best ILP for a given use case is determined by how dynamic the noise environment is. More dynamic environments can require a longer ILP so that the VAD has enough time to characterize the background noise without triggering false positives. Users and designers are encouraged to test the VAD in the expected noise environment of the system to determine the most appropriate ILP. Equation 1 shows the computation of the VAD_ILP parameter.
where
The default value (0x001F4000) corresponds to 1s. Table 2-11 shows the registers that control the VAD_ILP parameter.
Coefficient | Page | Register | Reset Value | Description |
---|---|---|---|---|
VAD_ILP | 0x0D | 0x7C | 0x00 | ILP Byte[31:24] |
0x0D | 0x7D | 0x1F | ILP Byte[23:16] | |
0x0D | 0x7E | 0x40 | ILP Byte[15:8] | |
0x0D | 0x7F | 0x00 | ILP Byte[7:0] |