The 'speech recognition' technology used in voice assistants converts sound waves into text. It generally operates on the cloud and leverages AI-powered models (e.g., deep learning). The microphone captures the voice and sends it to the voice assistant, which then understands the commands and generates a response. So, how reliable do you think these systems are? Or what causes the glitches?
How do voice assistants recognize voices?
👁️ 8 views💬 5 replies❤️ 0 likes
5 Replies
Voice recognition systems have come a long way, but sometimes they still don’t get my commands right—or they misinterpret them and give me hilariously wrong responses. How does yours handle things?
Thanks for the informative post! I was wondering how well these systems perform with different accents or in noisy environments, do you have any thoughts on that?
Last week, I tested Siri with my daughter and I can’t get that situation out of my head. When I told it, “The faucet is dripping, please fix it,” it didn’t respond at first. Then I gave the command, “Open the internet and find a faucet repair video.” This time, it opened an app on the screen called “Plumber”… Do you think they understand commands that well? 😅 From my experience, voice assistants are usually good with simple commands, but they mess up with complex sentences or when you speak with an accent. Since they run on the cloud, their response time slows down when the internet speed drops too.
It happened last week, I was at home in the afternoon. I didn't feel like ordering from the market, so I picked up my phone and said, "Alexa, order milk from Amazon." The first time, it didn't understand, so I repeated it loudly and clearly. I almost word-for-word repeated the phrase "order milk from Amazon," but instead of "Your order has been placed," I got "Sorry, I don't understand right now." In the end, I had to order manually, and I was pretty annoyed. Later, I found out that there was noise at certain frequencies in the audio file Alexa sent to translate in the background, so the model misinterpreted the command because the microphone couldn't capture the sound clearly. Strangely enough, when I tried the same phrase again in the morning, there were no issues. So it seems that noise levels, microphone sensitivity, and the quality of voice data processed in the cloud can instantly affect the system. It was such a simple command, but sometimes technical glitches can really frustrate you.
Ah, based on my firsthand experiences with speech recognition systems, you generally get about 90% accuracy when speaking clearly and distinctly. But the issues I run into most often are background noise or speaking too quickly. For example, lately, these systems haven’t recognized any of the commands I’ve given while my kid was making noise in the background at home.
A lot of the glitches mentioned in this thread can actually stem from the device’s hardware. The quality of the microphone directly impacts how clearly it picks up sound waves. If you try using it with cheap speakers or devices with low-quality mics, you’re likely to be disappointed. For instance, in my own experience, I had to pause for a second after saying “Okay Google” before giving a voice command, otherwise the words would get jumbled.