In the telecommunications industry, artificial intelligence already plays a lead role for marketers. For example, VoIP platforms are using IBM’s Watson to gather deep insights about callers and conversations. This data is then used to enhance and personalize the customer experience.
Inbound phone calls are a rich source of data for any company, and they are easier to mine and analyze with AI. For example, Watson transcribes conversations and converts them to text. The text can then be analyzed by software programmed to identify certain phrases and words designated as signals indicating actions. For example, the phrase, “what’s your warranty?” might indicate an intent to purchase.
With AI, large amounts of unstructured voice data are easier to analyze, but what about using it to improve the quality of those calls?
Analog audio doesn’t cause latency like digital does
For years, analog ruled the world of music and telecommunications. Then digital audio entered the scene and became the standard. Digital audio is easy to reproduce, but just like with analog the quality fades with each reproduction.
Sound quality of any kind naturally degrades over time and distance, as we’ve all experienced with cell phones, Skype, and other digital voice platforms. However, digital audio has the highest potential for latency because it’s not a complete signal. Digital sound consists of multiple points of data rather than one complete signal.
The difference between analog and digital audio communications
Analog phone lines—referred to as “landlines”—transfer voice data over a copper wire. Voices are converted to analog electronic signals and passed continuously in both directions. Analog signals are continuous and clear. VoIP converts voices to digital signals and sends them over a network. Digital signals are a collection of values that represent the original information. While digital voice can be clear, the data points can be lost or become corrupt in transit and that’s why digital voice is sometimes distorted.
When voice is transferred over a network, it takes more processing power to deliver the data. For example, with a landline, the copper wires and other infrastructure are already in place, and they were engineered to work with standard telephones. As long as the standard telephone doesn’t get too much of an upgrade, the infrastructure will support the calls. Digital communications are different because each person has his or her own equipment, which may not be of the highest quality. A person with top-of-the-line networking equipment won’t receive a clear digital voice communication from someone who sent it using poor technology.
With VoIP technology, businesses have been dealing with the effects of latency in calls for years. Latency occurs when the demand for processing data exceeds the capabilities of the system. For example, limitations in an audio recording system commonly create a delay in sound output. This creates inconsistencies in timing and errors in sound processing.
The same thing happens with voice calls, and it’s time for a change.
Reducing latency in VoIP communications with AI
During an audio conversation over a network, 300ms of latency will make people talk over each other, and the conversation will be impossible to follow. Latency isn’t just calculated one way. The time it takes to encode audio and the time it takes for the packet to travel are taken into account. Latency doesn’t affect audio quality, but it can make a conversation awkward.
During a voice call, every millisecond counts. “The typical round-trip VoIP flow constitutes a minimum of six packets, 20ms each,” explains Cato Networks. “That leaves just 130ms to 180ms of cushion for the entire transmission network. This may sound sufficient, the reality is that, after delays due to distance and network processing, that window is scarcely sufficient for delivering quality voice on a well-run network, let alone one with the variance of the public Internet.”
Clear voice calls rely on the successful delivery of packets, but errors can cause packets to be discarded. By detecting problems before packets are sent, AI can reduce the latency and loss of IP packets sent during voice communications. Detecting quality degradation early is the key to improving all the factors that affect voice calls: latency, jitter, and packet loss. Nothing can detect degradation as early as AI.
AI can also prioritize traffic more efficiently than traditional methods for routing network traffic and eliminate the need to stack protocols or create custom configurations that become complex sources of frustration later on.
What are we waiting for?
Three hundred milliseconds of latency today seems equivalent to waiting three weeks to receive a letter delivered by horse. Although, when the postal system was first created, three weeks was probably considered fast compared to what people were used to.
It’s time for AI to step in and speed things up
Keeping latency low is a significant task. Formerly, upgrading hardware or laying a new cable to create a more direct route were enough. Technology has caught up, and now processing speed is the main problem. For example, your connection might be low-latency, but if the network you’re connecting to has high latency, you’re going to feel it.
Network latency and application performance are one
Of all the ways to use AI, a boost in processing speed will bring the biggest benefit to telecommunications. AI has the power to store, move, and process data faster than current technology. With the aid of database management software powered by AI, it will be an unstoppable solution.
This article is published as part of the IDG Contributor Network. Want to Join?
