What is DTMF?
Also called touch tones, dual-tone multi-frequency, keypad input.
DTMF, short for dual-tone multi-frequency, is the signaling scheme that turns a telephone keypad press into a pair of simultaneous tones. Each key combines one tone from a low group and one from a high group, so a receiver can identify the digit unambiguously. It remains the standard way callers enter numbers and menu choices on a voice call.
The design uses two sine tones per key, chosen so that no tone is a harmonic of another and speech is unlikely to reproduce a valid pair by accident. A receiver detects the pair with narrowband filters or a Goertzel algorithm and reports the digit along with its duration. The keypad also defines star and pound keys plus a rarely used fourth column of letter keys.
Over IP telephony, digits can travel three ways. They can remain audible in the media stream as in-band tones, be sent as named telephone events in RTP as defined by RFC 4733 and its predecessor RFC 2833, or be signaled out of band with SIP INFO messages. Compressed low-bitrate codecs can distort in-band tones, which is why the RTP event method is generally preferred.
In voice applications DTMF is valued for precision. Account numbers, dates, and card digits are error-prone to transcribe from speech but exact when keyed, and keypad entry works when the caller is in a noisy place or does not want to say a value aloud. Menus often accept both speech and keypad input for the same choice.
For sensitive values, secure entry patterns suppress the tones from recordings and transcripts, or route digits to a separate collector so the values never enter the general call path. Handling card data this way is a common requirement of payment security standards, and the details of what may be recorded are set by those standards rather than by the telephony platform.
Key points
- Each key sends one low and one high tone simultaneously
- Carried in-band, as RTP telephone events, or via SIP INFO
- RTP events survive low-bitrate codecs better than audible tones
- Preferred for exact values and noisy or private settings
- Sensitive digits are suppressed from recordings and transcripts
In practice
A voice agent asks a caller to key in a six-digit reference rather than read it aloud, because the line is noisy. The caller presses the digits, which arrive as RTP telephone events rather than audible tones. The platform assembles them, echoes the reference back as speech for confirmation, and marks the event stream so the digits are excluded from the call recording and the stored transcript.