Court reaffirms validity of voice recognition
Lansing — A recent state appeals court decision has reaffirmed the validity of voice recognition to put a Michigan suspect in prison, even as critics contend it can be fallible evidence just like eyewitness testimony.
The Michigan Court of Appeals in a Nov. 14 ruling approved the use of a sister’s ear-witness testimony to help convict Detroiter Kelvin Junior Nolen of first-degree murder in the November 2014 shooting and killing of 26-year-old convenience store employee Mohamed Zokari on Detroit’s east side.
A court-appointed attorney for Nolen, now 29, argued that the Wayne County Circuit Court judge should not have allowed the sister’s testimony. A hoodie obscured the man’s identity on surveillance camera footage that was later used in court.
Nolen’s sister, Kenyatta Jones, testified that an audio recording of the man’s voice matched her memory of her brother’s voice. Jones also said the man in the surveillance footage had similar “physical characteristics and mannerisms” despite the fact that “she was not 100 percent certain of her identification,” according to the unanimous opinion signed last week by Court of Appeals Judges Jane Beckering, Colleen O’Brien and Thomas Cameron.
The appeals court affirmed that the Wayne County court was not mistaken when it allowed Wayne County Prosecutor Kym Worthy’s office to use Nolen’s sister’s testimony as evidence that ultimately led to his life conviction. He is incarcerated in the Michigan Reformatory in Ionia.
Calls to Nolen’s second court-appointed attorney, Jonathan Simon, were not returned. Evan Callanan, a Westland attorney who represented Nolen at the circuit court level, agreed the court should not have allowed the audio testimony. He said Nolen’s sister was not certain enough that it was her brother’s voice in the recording for the court to convict him beyond a reasonable doubt.
“I suffered after that loss, believe me,” Callanan said. “I believe he was innocent. I did the best I could for that guy.”
He also said a woman called two weeks before trial informing him that she was in the gas station 20 minutes before the shooting took place and that Kelvin was not there that morning. A judge issued an order to detain the witness at his request after she backed out of testifying, but police ultimately could not locate her, Callanan said.
The case is not remarkable in its admission of audio identification as evidence, said David Moran, a professor of law at the Michigan Innocence Clinic with the University of Michigan Law School, who is not involved in Nolen’s case. The clinic works to exonerate people who were wrongfully convicted.
“Like eyewitness identification testimony, it’s not 100 percent and mistakes are made,” Moran said. “But in one respect, it’s better than eyewitness identification testimony because (that) involves complete strangers trying to identify somebody.”
In this case, Nolen and his sister hadn’t been in touch for more than two years. But Nolen and Jones knew each other.
The two were adopted by separate foster parents as children but they lived in the same neighborhood and spent time together. They later went to the same middle and high schools for years, and Nolen lived with Jones for nearly a year between 2011 to 2012 as an adult.
“Because this evidence was sufficient to establish that defendant’s sister was sufficiently familiar with defendant’s voice to enable her to identify it, a proper foundation was established under (law),” the appeals court wrote.
“Further, given defendant’s sister’s familiarity with defendant’s voice, the fact that she had not spoken to defendant since the winter of 2012 does not compel a conclusion that she would not have been able to recognize his voice in April 2015.”
Moran said there’s a well-established practice of using such ear-witness testimony in making convictions in Michigan. And while it can be fallible, the jury also had other evidence, he said.
But just like convictions relying on eyewitness testimony, there have been well-documented mistakes.
David Shawn Pope served 15 years in prison after prosecutors in a 1986 Texas trial used “voice print analysis” to argue that messages left on the victim’s answering machine matched Pope’s voice, according to a group called the Innocence Project.
Pope was pardoned by Texas Gov. Rick Perry in 2001 and released from prison after the Dallas County District Attorney’s Office got an anonymous call that spurred the prosecution to run DNA tests on the rape kit showing that Pope was not the rapist. He had been sentenced to 45 years in prison.
The Michigan Court of Appeals argued that the circuit court did not make any error in admitting audio evidence in Nolen’s case. It was among a series of objections that Nolen and his attorney made about supposed mistakes in the trial that the appeals judges rejected.
“Viewed in a light most favorable to the prosecution, the evidence was sufficient to support defendant’s identity as the person who shot the victim,” according to the unanimous opinion. “At the time, only two people — the victim and a man in a black hoodie — were in the store. Thus, it was reasonable for the jury to infer that the man in the hoodie was the person who shot the victim.”
Nolen told the court he was at his girlfriend’s house from 6 p.m. on Nov. 3, 2014, until the following morning — past the time of the shooting, which took place Nov. 4 at 6:54 a.m. or 6:55 a.m., according to court documents.
But the prosecutor also used other evidence to question Nolen’s alibi.
Nolen’s cell phone data showed he called his girlfriend at 12:34 a.m., several hours before the shooting, the prosecutor said. Nolen would have had no reason to call her at that time if they had been together, the prosecutor argued.
Nolen demanded a new trial because he said the court improperly allowed the use of the cell phone records. The appeals judges rejected his argument because the rebuttal testimony “tended to contradict defendant’s evidence that he and his former girlfriend were together during the night preceding the shooting.”
Though UM’s Moran said the audio evidence in this case may have been justified, he acknowledged that in other cases it can be flawed.
“Like all forms of identification evidence, there’s a built-in error rate with audio identification,” he said. “The problem with all such identifications is that the human brain is not like a tape recorder, and so it doesn’t play back things exactly precisely, and the memory can be corrupted or suggested by any number of means.”