Nuts and Volts - September 2017

chip can be used to collect samples of audio levels at seven frequencies (Figure 10).

By toggling the Reset line and then pulsing the Strobe line a specific number of times, you can select the specific frequency you would like to measure (Figure 11).

Now that we have a chip that’s capable of capturing amplitude at multiple points in the spectrum, we need to connect this chip to an equally capable microcontroller. Steve chose the venerable PICAXE 20X2 chip.

By using its built-in Analog-to-Digital Converter (ADC), we can sequentially take samples from each frequency of interest.

Can You Hear Me Now?

When it comes to voice, frequencies below 400 Hz are not typically associated with jaw movement. Most are “plosives” associated with unvoiced sounds such as “t,” “k,” and “p,” and when recorded, are actually amplified as an artifact of the recording process (i.e., air expelled during speech that impacts the microphone element).

Alternately, frequencies above 2. 5 kHz are typically sibilant sounds such as “ssss”/“shh,” and the human mouth normally produces these by bringing the teeth and jaw together. This means the frequencies we can sample from the MSGEQ7 chip for jaw motion are 400 Hz, 1 kHz, and 2. 5 kHz.

The ADC in the PICAXE allows us to sample the amplitude on each of these frequencies with eight bits of resolution (value between 0 and 254).

So, rather than the original one bit of resolution offered by the Scary Terry circuit (i.e., ON and OFF) or the two bits of resolution offered by the Jawduino circuit (i.e., OFF, LOW, MED, HIGH), this new design will allow us to sample amplitude with eight bits of resolution (255 discrete positions) across three distinct frequency bands — quite an improvement!

If we sample each of these, we can get a clear picture of when the jaw needs to be open in order to more closely model the jaw position in relation to an audio file.