United Arab Emr
Asked — Edited

Multichannel For Voice Commands With ARC

I am looking to use ARC voice commands within the house . However, I an not sure how to make ARC monitor multiple microphones/channels or what audio system needed to facilitate my request from any room . Any ideas ? :D


ARC Pro

Upgrade to ARC Pro

Stay at the forefront of robot programming innovation with ARC Pro, ensuring your robot is always equipped with the latest advancements.

United Kingdom
#1  

Parallel your mics on the same input but be aware that it may increase unwanted noise on the mic.

Ireland
#2  

I would suggest you have two options depending on your layout 1 : A clip on radio or bluetooth mic 2 : A number of microphones connected into a mixer unit.

#3  

I would say a mixer unit of some kind. if usb mic.s, find a program that mixes the audio from them all to one channel.

United Kingdom
#4  

The mixer would need to also provide a windows device selectable within the speech recognition settings. An external, standalone mixer would be better, with it's output fed in to the mic or line in of the PC sound card.

Other Jarvis project I have seen while searching for bits and bobs and ideas for mine have simply paralleled the mics and claim to have good results. It's a cheaper alternative but you lose any individual control over the mics.

United Arab Emr
#5  

Thx all for your quick response . I was hoping that someone witin the group already gone into this track . Probably, I have to go for the expensive option using a mixer with wireless mic to minimize the wiring in the house and maintain the audio quality. I will go into more investigation for this scenario.

United Arab Emr
#6  

@rich: it seems that you provided new Feedback while writing mine I understood from your Jarvis comment that audio quality is critical for the commands recognition That is why, I am concerned about audio quality and same time wiring .

United Kingdom
#7  

How many people are to use it?

It's something I've looked in to a lot for Jarvis since he is everywhere. Currently I use a bluetooth handsfree headset which gives me the ability to talk to him from where ever I am in the house but it is limited to just me.

I did try a few ideas out with multiple mics dotted around the house but the problem with that is the SR software picks up everything and had too many false matches for my liking. Also, when it was noisy it was difficult for it to understand what to listen to and what not to.

Ideally, wireless mics on anyone who needs to talk to it, preferably with a button which enables it to minimise background noise and idle chatter that interferes with the listening.

@Toymaker has posted previously about the mic he uses, while not cheap it is a very good mic. I think it was in one of his topics but I really can't be sure.

United Arab Emr
#8  

Ultimately, the idea is to use the system by the family and to be triggered by a specific words . Depending on the solution complexity hoping to detect the room where the commanding voice coming from.

On the other hand , what you are suggesting is more practical to implement with less cost and efforts.

United Kingdom
#9  

Jarvis is triggered by the word Jarvis. I don't use ARC for the VR since I have it all set up in a different program and it controls more than just ARC. Even with a confidence of 90 on the pre-word "Jarvis" and complex(ish) commands such as "what is the weather like" with a confidence of 80 it still picks background noise up as these commands... Earlier today I was watching TV and all of a sudden it triggered "Jarvis I am bored entertain me" (my headset is on charge close to the centre speaker which is why it heard the TV).

Basically what I am saying is, even with a "magic word" it can still give false commands from background noise. Increasing confidence means a decrease in successful commands, reducing it can result in false commands. It's an issue that needs balancing out and one I'm still looking for hands free solutions for.

If you want room detection you will need each mic to be addressable somehow, through a software mixer for instance. Or, by using PIR detectors in each room to detect where the activity is - not ideal for multiple people though.

United Arab Emr
#10  

Thx rich for your feedback and sharing your experience.

#11  

I have the HAL2000 software from Home Automation Living in my house that controls everthing from lights, appliances, heating and AC. I also have several mics through the house and a speaker in each room so I can hear HALs response, for example "Computer what is the weather" "the weather is sunny". The speaker also allows me to hear the return response from HAL just in case he did not understand, example "Computer what is the weather" "you want me to turn on the TV""No, what is the weather". One big issue I had was with multiple mics was the back ground noise from all rooms triggering HAL. I solved this over time by adding 2 motion detectors in each room, one for the mic and one for the speaker. When I am not in a room the circuit to the mic and speaker are off so the system cannot pickup any background noise and when HAL responds to me only that room will address me. This works well unless someone is in another room they will here HALs response, but for the most part it does work better than having all the rooms listening and speaking.

United Arab Emr
#12  

@Tymtravler: How did you connect the mics? In parallel ? It does not affect the audio quality and recognition ? And yes, the PIR is good if you are alone in the house .