This project is for an audio engineer to evaluate, test and recommend microphone amplifier and AGC settings using the Maxim 9860 Audio Codec. An evaluation kit is available ($63.73 at DigiKey) for testing and concluding recommendations.
The objective of the effort is to recommend settings for gain and ACG parameters so that our Speech Recognition Accuracy is optimized in different noise environments. Proper alignment of the AGC window with optimal attack, decay, hold-times, release times and thresholds is expected for a variety of headsets and background noise environments. You will need to optimize using 8x8 uLaw companding.
The goal is to produce a table of recommended settings for three typical commercial noise environments (think of a Chucky Cheese restaurant before lunch, during lunch and with a kids birthday party under way). This table is then repeated for three headset variations (headsets provided). Headsets will have either a short boom mic or an in-cable mic.
The gain settings and AGC window should be aligned for maximum speaker dynamic range in each of the three noise environments. Calibrated sound pressure levels for the average noise should be included. Average RMS values of the optimum WAV files should be included. We will be looking for recommendations for the following parameters on the Maxim codec (current setting are shown):
amixer set 'Master' '186'
amixer set 'Master Playback Boost' '0'
amixer set 'Sidetone' '0'
amixer set 'Mic PGA' '20'
amixer set 'Mic Pre Amp' '+20'
amixer set 'Mic1' '15'
amixer set 'Mic2' '15'
amixer set 'Playback Function' 'None'
amixer set 'Capture Function' 'None'
amixer set 'AGC Capture Attack Time' '12ms'
amixer set 'AGC Capture Function' 'Left output'
amixer set 'AGC Capture Hold Time' '400ms'
amixer set 'AGC Capture NG Threshold' -- '-28'
amixer set 'AGC Capture Release Time' '1.25s'
amixer set 'AGC Capture Signal Threshold' '10'
Timeframe These results are needed in January at the latest As always, the sooner the better. Please be specific about how quickly you can get to this task and your estimates of how long it will take to complete.
I have BS, MS and PhD degrees in Electrical and Electronics Engineering from one of the most reputable universities in Turkey and have 15 years of analog and microcontroller based systems design experience (both hardware and firmware) and 5 years of RF design experience. I am founder and manager of an R&D company focusing on analog and embedded systems design.
I have a high level of audio circuit design and codec management experience. Please visit my profile, where a Class-1 SPL measurement tool that has passed the related anechoic room tests is presented. I have gone through the MAX9860 datasheet and eval. kit datasheet. I can complete the task in about 20 days once I get the kit and headsets. However, getting the headset through DigiKey can take 15 days considering the new year holidays and in total, the project may extend to the beginning of February. For optimization, I can use .wav files provided by you as well as my recordings or I can even test on the field.
Please note that the requested prepayment is only for nulling the immediately taken 10% site commission. I found it necessary because of employers who disappear after awarding the project. It can be replaced by an early milestone.
For best results the audio input optimization will also take into account the front-end type/parameters of the speech recognition engine that follows. After briefly glancing over the chip's specs I find it puzzling why you compromise the quality from the start with the uLaw companding. This is perhaps due to the larger picture so if you have an interest in my bid we can clarify such issues latter...
Kind regards,
x3x7