Challenges and Benefits of Speech Recognition and Voice Control
The concept of speech recognition is not new. The use of speech recognition and voice control has been researched for more than two decades. With the advancement in computer hardware and software, the voice control applications have become more practical, reliable and flexible. But, the use of speech recognition and voice control in the flight-deck requires them to be highly capable and reliable which the current speech recognition and voice control systems have not reached yet.
If you use SIRI (Iphone) or CORTANA (windows 10) more often then you may think that if your phone misspells many times your dictation how an aircraft with voice activated cockpit (VAC) could be safe? The communication with the cockpit should be as reliable, flexible as man to man communication. The current speech recognition accuracy rate falls in the range of lower to mid 90%. Some speaker-dependent systems or applications with small vocabulary of commands may show accuracy rate in the upper 90% range. Even with this higher accuracy rate of speech recognition, it is not reliable because out of every 10 words 1 word will be incorrectly recognized. Under ideal conditions, higher accuracy rates can be achieved under controlled lab environments but under actual operating conditions different technical challenges arise which should be overcome.
Some of the issues presented by an actual operating environment are listed as follows:
- High noise levels including cockpit noise, random noises, bumps and thumps, multiple people talking at once
- A range of variation of operators’ accent and the need to train the system to recognize the operator’s voice patterns
- Changes in the speaker’s voice due to illnesses, air pressure, vibration, G forces
- Limited command sets because of the high level of hardware resources previously required to process speech
- Necessity to memorize and speak a limited set of commands
- Necessity to speak slowly, one word at a time
- The inability to differentiate between words or phrases intended as a command and those that are part of a conversation
The above listed problems have been resolved quite significantly in the latest generation automatic speech recognition applications. Lockheed Martin’s F-35 Joint Strike Fighter (JSF) uses Automated Speech Recognition System which surprisingly meets the performance requirements. Also, Speech recognition and voice control system has been successfully tested on UK ARMY Air Corps Gazelle helicopter. The system is speaker independent and does not require any user to train the system before using the system. It gives aircrew the ability to control aircraft systems using voice commands and access information without removing their hands from the flight controls or being distracted from what’s happening outside the aircraft. But using speech recognition and voice control system in a commercial airliner still poses a greater risk because of requirement of very high level of reliability and capability of the system.
If such is the case why use a voice control in the flight-deck? Here are the benefits of using speech recognition and voice control systems. Under stressful and emergency conditions, pilot’s visual channel is maxed up while manual channel remains moderately loaded and typing long strings of alphanumeric, turning knobs and dials are susceptible to be incorrect since they require pilot’s close attention. But, his auditory channel which is used to communicate with ATC, copilot and passengers is relatively lightly loaded and this could be exploited to take its advantage for controlling the aircraft. A complicated command from ATC can take pilots up to 30 seconds to turn all the knobs, hit all the buttons and make that change occur. The speech recognition system cuts that time to eight seconds giving pilots 22 more seconds to spend keeping the airplane safe.
Following are some examples of safety and efficiency benefits of speech recognition system:
- The pilot can just ask what he needs to know about without stepping through menus of specific system, scanning the dials. For example, he can simply ask “say remaining fuel” to know the state of fuel.
- The data entry of FMS, autopilot and radio frequencies become much easier, faster and safer since the speaking incorrect input data is less likely.
- The system correlates unfamiliar local data with ease. For example, on situations like you know the place you go to but PFD shows abbreviation seemingly totally unrelated then you just have to repeat the place and the SR system will do rest.
- The pilot may be able to configure the glass cockpit quickly to his or her preference by simply announcing.
- A suitable VAC could monitor the pilot’s read-back of assigned ATC headings, altitudes, and altimeter instructions and compare the read-back against both what it heard ATC say as well as what the aircraft systems are reporting.
Article By: Govind Kumar Chaudhary (2069/BME/416) THAPATHALI CAMPUS