Enhance Audio Processing Pipeline with Parameterization and RMS Calculation

[Carpediem324]

This pull request introduces two key improvements:

1. streams.py Update:

   • Added channel and audio rate parameters to support flexible audio input configurations.
   • These changes resolve errors caused by mismatches between the user's microphone settings and the engine's requirements.
   • Implemented channel mixing and downsampling to convert the incoming audio to a single channel at 16,000 Hz.

2. engine.py Update:

   • Enhanced the scoreFrame() function to compute the RMS value of the input audio frame.
   • The calculated RMS value is now included in the detection result, providing an additional diagnostic metric for audio signal intensity.

These updates improve the robustness and compatibility of the audio processing pipeline, ensuring that audio from various sources can be seamlessly integrated with the engine's processing requirements.

[TheSeriousProgrammer]

Looks good will review it once I reach home

[TheSeriousProgrammer]

My bad thought the resampling and channel averaging is taken care automatically by pyaudio, the mic with which I test worked seamlessly, after a quick search realized that it's because of the audio driver on its own supporting the same l, this may not be the case with all

[Carpediem324]

While testing, I discovered that mismatched sample rates resulted in errors, so I updated the code accordingly. Furthermore, based on the idea that RMS helps differentiate keywords across multiple devices, I incorporated RMS into this revision. Thank you.

[TheSeriousProgrammer]

Also can you give an example of how you use the rms value

[Carpediem324]

Using RMS (Root Mean Square) in keyword detection provides two main benefits:

First, in a multi-device environment, RMS helps select which device should respond when multiple devices detect the same keyword simultaneously. The device with the highest RMS is typically closest to the user.

Second, even in single-device scenarios, RMS helps filter out false detections. If the RMS value of detected audio is too low, it likely indicates background noise or distant sounds, allowing developers to avoid unintended activations.